Audrey Le Gouellec

Dr Audrey Le Gouellec is an Associate Professor and Hospital Practitioner in clinical biochemistry at the Faculty of Medicine, Université Grenoble Alpes and Centre Hospitalier Universitaire de Grenoble Alpes.
Learn more about her work at CHU Grenoble.
Discover the activities of the French Speaking Network for Metabolomics and Fluxomics (RFMF).

Favorite metabolite
Spermidine & hypusine

Papers discussed in this episode
Metabotypes of Pseudomonas aeruginosa Correlate with Antibiotic Resistance, Virulence and Clinical Outcome in Cystic Fibrosis Chronic Infections
Moyne et al. 2021 Metabolites https://doi.org/10.3390/metabo11020063

What clinical metabolomics will bring to the medicine of tomorrow

Le Gouellec et al. 2023 Mini review Frontiers in Analytical Science https://doi.org/10.3389/frans.2023.1142606
More about the Fibrotest discussed in this episode.

Sign up for The Metabolomist mailing list to be the first to hear about the latest episodes and news around metabolomics at https://themetabolomist.com

Finally available – Alice’s first book
The STORY principle – A guide to the biological interpretation of metabolomics
Also featuring some of the Metabolomists from Season 1
Available on Amazon and the biocrates webshop

Sandi Azab

Sandi M Azab is Research Associate at McMaster University Department of Health Research Methods, Evidence and Impact in Canada.
Learn more about her work here. Connect with her here.

Favorite metabolite
Pentadecanoic acid (FA 15:0 or C15:0)

Papers discussed in this episode
Early sex-dependent differences in metabolic profiles of overweight and adiposity in young children: a cross-sectional analysis Sandi M Azab et al. BMC Med. 2023 | https://bmcmedicine.biomedcentral.com/articles/10.1186/s12916-023-02886-8

Systematic review discussed in the episode

A systematic review of metabolomic studies of childhood obesity: State of the evidence for metabolic determinants and consequences
Evangelos Handakas et al. Obesity Reviews. 2022 | https://onlinelibrary.wiley.com/doi/10.1111/obr.13384

Sign up for The Metabolomist mailing list to be the first to hear about the latest episodes and news around metabolomics at https://themetabolomist.com

Finally available – Alice’s first book
The STORY principle – A guide to the biological interpretation of metabolomics
Also featuring some of the Metabolomists from Season 1
Available on Amazon and the biocrates webshop

Marc-Emmanuel Dumas

Marc-Emmanuel Dumas is Chair in Systems Medicine at the Department of Metabolism, Digestion and Reproduction of the Faculty of Medicine at Imperial College London in the UK and CNRS Director of Research and group leader at U1283/UMR8199 at the European Genomic Institute for Diabetes in Lille, France.

Learn more about the work of his team here.

Favorite metabolite
Phenylacetic acid (PAA)

Papers discussed in this episode
Molecular phenomics and metagenomics of hepatic steatosis in non-diabetic obese women
Lesley Hoyles, José-Manuel Fernández-Real, Massimo Federici et al. Nature Medicine. 2018 | https://doi.org/10.1038/s41591-018-0061-3

Sign up for The Metabolomist mailing list to be the first to hear about the latest episodes and news around metabolomics at https://themetabolomist.com

Finally available – Alice’s first book
The STORY principle – A guide to the biological interpretation of metabolomics
Also featuring some of the Metabolomists from Season 1
Available on Amazon and the biocrates webshop

Vadim Gladyshev

Vadim Gladyshev is a Professor of Medicine at Brigham and Women’s Hospital, Harvard Medical School
Discover more about the work of his team here.

Favorite metabolite
Selenocysteine

Papers discussed in this episode
DNA methylation, transcriptomics and metabolomics to study biological age after surgery, in pregnancy, and severe COVID-19.
Biological age is increased by stress and restored upon recovery
Jesse R Poganik, Bohan Zhang, Gurpreet S Baht, Alexander Tyshkovskiy, Amy Deik, Csaba Kerepesi, Sun Hee Yim, Ake T Lu, Amin Haghani, Tong Gong, Anna M Hedman, Ellika Andolf, Göran Pershagen, Catarina Almqvist, Clary B Clish, Steve Horvath, James P White, Vadim N Gladyshev. Cell Metabolism. 2023 | https://doi.org/10.1016/j.cmet.2023.03.015

Metabolomics
Organization of the Mammalian Metabolome according to Organ Function, Lineage Specialization, and Longevity
Siming Ma, Sun Hee Yim, Sang-Goo Lee, Eun Bae Kim, Sang-Rae Lee, Kyu-Tae Chang, Rochelle Buffenstein, Kaitlyn N Lewis, Thomas J Park, Richard A Miller, Clary B Clish, Vadim N Gladyshev. Cell Metabolism. 2015 | https://doi.org/10.1016/j.cmet.2015.07.005

Sign up for The Metabolomist mailing list to be the first to hear about the latest episodes and news around metabolomics at https://themetabolomist.com

Finally available – Alice’s first book
The STORY principle – A guide to the biological interpretation of metabolomics
Also featuring some of the Metabolomists from Season 1
Available on Amazon and the biocrates webshop

David Wishard

Wishart group @UAlberta

Favorite metabolite
tryptophan

Papers we discussed in this episode

• Emerging applications of metabolomics in drug discovery and precision medicine
• Towards a Rosetta stone for metabolomics: recommendations to overcome inconsistent metabolite nomenclature

Other resources
David Wishart’s talk at the event on Pan-cohort studies in October 2022
Why targeted metabolomics is essential for population health

Sign up for The Metabolomist mailing list to be the first to hear about the latest episodes and news around metabolomics at The Metabolomist

Episode Transcript

Alice: Today, I’m joined by David Wishard. You are a professor at the University of Alberta, and you are a very important member of the metabolomics community. You’re involved in many different parts of the life of metabolomics. I let you explain a bit your background and how you got to work with metabolomics and what are your favorite activities that you have at the moment in the field.

David: Sure. It’s, a long story. I got involved in metabolomics because I was trying to come up with a lecture topic about NMR that would be relevant to students in the pharmacy school. In pharmacy, the focus is on small molecules. And for most of my life I’ve been working in large molecules like proteins, and I used NMR to study proteins. So I was forced to do some quick research about small molecule NMR and medical applications in general.

In the course of doing the lecture and the course of building up the material for that, I realized that you could actually use some computers to simply solve a very challenging problem at the time, which is how do you deconvolute mixtures in an NMR spectrum. So at the time that I was doing this back in 1997, there wasn’t even a name for metabolomics: It wasn’t a term what we were doing. We just said, “that’s NMR”. At the time the term “metabonomics” started appearing, maybe in 1998. In 1999, I think the first paper on metabolomics, was called, we were calling it, Chenomics – So Chemistry and Omics. And that actually led to a little spinout company that’s still going (Chenomx). It does NMR based metabolomics. That’s how it got started. That’s how I got started in the area trying to find a topic to talk to students about with small molecules. But it combined my interest in programming and computing along with spectroscopy. Over the years I had to learn how to do GCMS and LCMS and I learned a lot of analytical chemistry that I never really took in school. I guess I’m a backdoor chemist. I came in the wrong way.

Alice: Your group is involved in many different sides of the work in metabolomics. You talked about the measurement, but you’re also very active in cataloguing metabolites. We will get to this later in the podcast. So with databases but also bioinformatic tools you’ve contributed creating and sharing a lot of tools with the community. So are there specific activities that you have or that your group has that you maybe want to mention now? – the main places where your group is visible in the community at the moment.

David: Sure. We have a mixed lab. The dry lab is the computational part, and the wet lab is people running the instruments but also doing sample prep, cell culture, or bacterial work or sample collection. You have molecular biologists, analytical chemists, we’ve got computer programmers, we even have an engineering team that helps with fabricating and designing things. It’s a real mixed group – And of them are working to solve each other’s problems because we all have problems to solve. Our dry lab does a lot of database development and a lot of software development to make untargeted metabolomics a little easier, but also to inform the community about what are involved in metabolites. Our wet lab group works primarily on targeted metabolomics and that’s something I’ve always believed is important because it’s the best way to quantify and I come from a background in NMR where we always quantify. It struck me as very odd that a lot of people in the metabolomics community weren’t too concerned about measuring concentrations.

I think, again, that is coming through viewing metabolomics from a different perspective. Something that we recognized early on was that there wasn’t a good collection of information about metabolites. And so that led us to establish these databases. The Human Metabolome Database (HMDB) was one, and then to develop software to help with analysis, and that was MetaboAnalyst.

And we wanted to do this on the web to make it really accessible to people. So both of those have done very well. We’ve created other resources I think that have also helped standardize things in the community, help inform people in the community. And I think that’s been a real theme for our work to try and democratize metabolomics, make it accessible, make it more amenable to people.

From the Instrumental and wet bench side we run a service facility now. It’s become a national facility in Canada – The so called the Metabolomics Innovation Center (TMIC) and that, too, is intended to democratize metabolomics making people and their research accessible to metabolomics resources because it’s an expensive business to get into.

Alice: This is also something we mentioned in the first season of the podcast – Metabolomics is really a multi-expertise kind of technique. You need to be good at completely different topics to be able to run a whole metabolomic study from beginning to end. And so I guess for someone who is interested in metabolomics but doesn’t have the resources or is beginning, it’s good to have this kind of facilities like TMIC, for example, where all this expertise is already there for you and you don’t have to be an expert at everything to begin with.

David: That’s right. It takes the village, as you say, it is very multidisciplinary. No one person I think knows all of the analytical methods. Chromatography, GCMS, LCMS, ICPMS, NMR. It’s a lot to learn and most people never get that experience in a single PhD or two or three PhDs. It helps to have these core facilities.

Alice: Do you also help with the interpretations of the biological interpretation? Do you have people working exclusively on this as well?

David: We do. And I think this is one of the biggest challenges in metabolomics. It’s getting easier to collect the data but to interpret or process the data is more challenging. The role in developing MetaboAnalyst was to help make the statistical analysis more easier but that’s not what you get really in terms of the biological interpretation.

It requires reading it requires a better understanding. We’ve been trying to develop pathway resources that would help with that. Trying to develop biomarker resources that would also help with the interpretation. And I think that the tendency in a lot of metabolomic studies is to sort of just stop at the statistical end and say, “I am done.”, and not really explore the biological side. And that’s where the really interesting things happen.

Alice: I agree. I don’t know if you heard about the story principle. It’s a book that I’ve been working on. In that book, I talk about like five steps to get people to the interpretation and exactly for what you said that – It doesn’t stop at the statistics, it doesn’t stop at the list of what goes up and what goes down and to really understand what’s going on or to make use of the metabolomics, you have to get to that e extra step to understanding the biology.

I’m also very interested in by informatic tools that help us get closer. It never gets exactly completely to the end story, but it can bring larger bricks of the house that you try to build when you build the story.

David: That’s right. I think that’s a really good point. Science is storytelling. Writing a journal article is a story. I think the other part to which is even beyond the interpretation and this is something that we’ve seen particularly in the clinic or applications to consumer-based systems. You tell them the story and then they say, “Well – now what? What do I do?”. And so it’s not just trying to explain what’s going on, it’s also trying to say how do you fix it? And that’s another challenge I think, which is perhaps even more compelling.

Alice: About your experience with metabolomics. I was wondering, is there something that you wish you had known before you started working with metabolomics that maybe you learned over 20 more years working with it? That’s, if you’d only known that before. Not necessarily something that we’ve discovered with new technologies, but something that as a beginner you don’t know, and then you figure it out later. And that maybe people could benefit from if they are not starting their work with metabolomics.

David: Yeah, when we originally started it was in 2006 it was called the Human Metabolism Project, which led to the development of the human metabolism database. At the time we started the project, we thought there were around 600 metabolites in the human body, and that was what was listed on all the encyclopedias and books. So we thought, this is going be pretty simple. At most we might get a thousand – in the first year we already were up to 2000. Now, some 17 years later, we’re about 250,000 – and that apparently only covers 5% of the true or known metabolism. Multiply that by another 20 fold so we’re maybe about 5 million compounds. I wish I’d known that it was going to be so big. The metabolome universe is so much larger than say the protein universe or the genome universe.
Alice: That is an interesting point. Then what would you say to people who are beginning with this and they’re thinking, okay, how many metabolites can I measure in one study, whether it’s targeted or un targeted, like it’s going to be maximum a few thousand, if I’m lucky. How is that going be relevant to the gigantic-ness of the metabolome? Where should I do this?

David: I think there two extremes that we’re looking at. I mean, one is to try and identify everything and assuming everything is important, but that´s like looking at a lawn: Is every blade of grass important or is the fact that the lawn is green or brown important. And I think as scientists, we’re often curious about the detail, but you still want to be able to see the broader picture. I think one of the things that’s emerged over the last 10 years is that there’s a common set of metabolites that are changed in disease. There are obviously some compounds, particularly compounds that we call the exposome that are causative for disease. And some of these are remarkably low levels. And so there’s a compelling case to be made to say, yes, we need to measure these obscure things that we didn’t know were there because maybe in fact they’re causing large numbers of conditions.

So I can see it from both sides, but from the perspective, how is it impacting on the body or physiology? Maybe there’s only about 400-500 key actors that we need to really look at and measure. I would emphasize that we need to measure those accurately and quantitatively. But it might even be that there’s a smaller set that we can work with to see what has changed and how it’s changed. If we want to understand what is causing the change, then in fact maybe this desire to measure everything will be important. Time will tell.

Alice: What you mentioned about using quantitative measurements is also really important in the context we are interested in today – The application in the clinics. You had a talk last year that is also on the biocrates YouTube channel where you make a strong point that these measurements should be quantitative if we want to have a chance to have them be useful in the clinics. We will come back to this at the end of the episode. For multiple applications this is something that makes a huge difference.

Let us discuss a bit more the HMDB and the different databases that you’ve been building over the years: HMDB comes from the Human Metabolome Project where you had a number of groups that were working together on first identified metabolites and then cataloguing them and then you ended up having something much bigger than you expected. This is always interesting because also when we use omics, we want to look at everything.

This was the same for all omics. And then, I guess, there was a shock at some point when you figured out. What you set out to do? Do you have any regrets?

David: Well, it keeps you busy. At least you’re still employed, when you have a bigger project than you expected. But it, yeah, it’s sort of, you start off with a meal thinking it’s only a single course dinner and find out it’s a 12 course dinner. When we started the human metabolome project, we thought it was more contained, more constrained, and that it was solvable and reachable. But I remember going to a meeting, I think it was in North Carolina, where someone highlighted the fact that metabolomics is probably more complicated than we expected, in part because we eat other metabolomes and and a light bulb kind of went off in my head and said oh – That’s very true. Because in fact we eat plants and plants have a very different metabolome than humans. And we eat a variety of other prepared foods. And so these things have chemicals added to them. Just a matter of doing a quick Google search and realizing that there are 5,000 compounds added to foods and the list of plant phytochemicals was over 300,000 listed in the natural products databases. And I just had a sinking feeling that we are going to be very busy for a very long time.
The intent of the Human Metabolome Project and the database itself is to try and capture that information, make it more accessible make it usable from the perspective of the metabolomics community. Make it searchable and give a standard hub the same way that genebank has helped the genetics and molecular biology; in the same way that the protein data bank has helped with structural biology.

Alice: Does it have to do with the nomenclature; with having a standardized way of naming things and describing things is – or what is for you the added value of having databases that collect everything in one place.

David: I think part of it is standardization. We spend a lot of time identifying the different ways of naming chemicals. Most of them have a dozen different synonyms. I think we also wanted to consolidate information that was scattered. The centralization is a way of at least getting that all in one place so you can look at it. Moving it to the web meant that you didn’t have to have a big book published. If we had it as a book it would be 50,000 pages if that – making it web accessible makes it more practical. It also makes it more searchable. We can search not only by names but you can also search by structures and that’s very easy, especially if you’re a chemist. You think in terms of structure. It also allowed us to put in diagrams like pathways it allowed us to put in spectra so that people could compare things. The visual data has become increasingly important in the database. I think there are other evolutions, revolutions that might be happening. The development of natural language processing and chatbots. It might be a way of allowing more specific queries to the database and giving you textual answers rather than read and read and read.

Alice: That would be great.

David: There’s certainly a way of evolving databases so that they are topic specific more interactive. And I think that’s one of the things that we’ll be trying with our databases over the next year or so.

Alice: Are there uses that have been made of those databases that surprised you? Things you didn’t expect?

David: Yeah, I think there’s always been a lot of surprises. Originally we started making a list not only of metabolites but we also wanted to track the drugs and then we also wanted to start tracking the foods. And so one database was called drug bank and the other one was Food db, and then another one is, human metabolism database. The drug bank, which was a small database became incredibly popular. And we didn’t know why. And then we found out that what people were most interested in was the drug target information that we put in because that hadn’t been put in in any data resource and people were using it to identify new drug targets and to repurpose drugs.

And so that was, “oh, I didn’t know you could do that”. But that’s grown into a very big resource for the drug industry. HMDB – a lot of people have used it for applications and compound classification; reinterpreting mass spectra in ways that we never expected; defining what is biologically or drug relevant or a natural product resource, or what is a natural product. We’re surprised at the ideas that people come up with with the databases. And I think that just sort of reinforces the need for putting it out there and letting people discover new ways of interpreting the data.

Alice: Recently you published with other co-authors a comment in nature metabolism. Saying you are going towards a Rosetta Stone for metabolomics with the recommendations to overcome inconsistent metabolite nomenclatures. Do you have a few key points that maybe you could tell us about, because this is also something we discussed on the podcast a couple of times. It can be difficult to make sure that you exploit your metabolomics to the fullest when you’re not sure that you’re actually naming your metabolites in a way that is understood by all the tools that you want to use. So what were the main points that you wanted to discuss in that paper?

David: I think, one of the more striking things that submerged is that there are several cases where people have rediscovered the same metabolite – over multiple years …and existed under multiple names and so people didn’t realize these things had been around or were discovered for certain applications or people were making claims that they were the first to discover it only to find out years afterwards that someone else had rediscovered it many years before so this is a problem and it’s a problem with nomenclature. It’s a problem how people use traditional names in the literature. There are solutions to using names and some of these are using things like inchy keys or standard unique identifiers. You could also use identifiers from databases, you can do structure searches to see if your molecule resembles something else that has already been known. Making those things available like a Rosetta Stone to help with even the translation of a compound name to what it really is or a chemical translator would help. I still think it’s critical for the journals to adopt this idea of using standard identifiers when compounds are listed or named. And that way we’re all on the same page and speaking the same language. I guess what some level the Rosetta Stone did when we were trying to convert hieroglyphs into the language we could understand.

Alice: To conclude on the topic of the databases: How could someone contribute to, for example, to HMDB? Do you take inputs from the outside and someone who’s interested, who maybe is doing a lot of identification of peaks or also maybe writing their PhD is on this metabolite. Then they have so many interesting literature references about it on functions and stuff. Do you take input from the outside and how does that work?

David: We do, we’d like to have more or we’d like to hear more from people about either corrections, they’d suggest or additions that they’d like to see and improvements that could be done. We have more by accident than by design started taking submissions on metabolites. So we’ve been bringing in mass spectra depositions, NMR spectral depositions, and more recently infrared spectroscopy depositions.

Alice: People write directly to the database, or how does that work?

David: Essentially they will write to us and say, can we deposit it? And then we’ll kind of busily work away. We have been developing deposition tools for a different project, but I think we’ve realized that those deposition tools would work just fine for the HMDB. So we’re thinking that it should or could be possible for people to deposit new compounds or new compound ideas and spectra associated with those compounds. That’s something that may be coming in the next year or so and then that may open the opportunity for people to build on the database or contribute to the database as scientists.

Alice: Great. Another topic I’m really interested in and that we’ve also talked a lot on the podcast is different bioinformatic tools to analyze and understand metabolomics. Of course we can talk about MetaboAnalyst. It’s always the first tool that I recommend to people who begin because it has a lot in one place. So it’s a great place to start. So do you want say a few words about MetaboAnalyst and maybe you want to tell us if you favorite bioinformatic tools. What are your favorite ways to take data and bring it to the biological understanding?

David: MetaboAnalyst is my go-to tool as well. It’s really easy to use and it’s great and it’s continued to be developed by Jeff Shaw over at McGill now. It is always nice to have students start out in your lab and if they really enjoy their work, then they can kind of move on with that and take that ownership.

I think the intention of MetaboAnalyst is to help you with the statistical analysis. And there’s a bit of biological interpretation it can provide you with, but not, quite to the degree that I think people need. Over the last few years we have been working on something called Path Bank or Small Molecule Pathway Database. And what we’re trying to do is capture more information, relating to the physiological effects, the association with metabolites in pathways to their proteins and enzymes, cells and organs and organelles that they’re located in and to help extend the biological interpretation beyond just simply – “this one’s up and this one’s down”.

There’s real utility in MetaboAnalyst for biomarker identification. But again, that’s an end in itself. If you want to understand the biology behind those biomarkers, again, it still requires moving towards pathways. Now pathways can only take you so far. As a rule, when I’m noticing metabolite changes and we’ll look at maybe some pathways I usually (still rather) use the literature. The challenge these days is that the literature is so vast that it’s hard to find the things you need. We’ve been working on an ontology for metabolomics, it’s called ChemFont and we’re using both the data in our databases as well as manual annotation as well as natural language processing to expand what’s in ChemFont. Ontologies are used by people and machines and they’re machine readable and the gene ontology in molecular biology has made a huge difference. So we wanted to have an ontology for metabolites and chemistry and this would allow people to do more meaningful interpretation of their data. The idea is to have an ontology with what we call triples, an object, a verb, and a subject that gives you information, “X does this to Y” or “A does this to C or comes from C”. By generating these facts or statements with references, it would hopefully allow people to save a lot of time that they’re reading the literature, but then they can also start synthesizing ideas or using the computer to help synthesize ideas. And so the intent of having this ontology, combining it with smart language tools or large language models like chatGPT hopefully would give people the ability to extract more useful data in the biology about their metabolite sets.

Alice: And so getting every time a little bit closer to what the human can do. Yes. So that the human has less work to do that could be automated. It’s always the dream. Especially when there’s so much literature now that you can have a tool that helps you to cut through the weeds and then extract the beautiful things that pre-read it for you, and then you can do the final step yourself. But you always have to do the final step yourself, though.

David: I think ultimately you still have to weight: I’m seeing three references that say this and two references that say the opposite. What do I need to do? Exactly What does it mean?

Alice: You probably answered that question just now. What are there tools that you wish existed? Maybe this is one of them? …and it is about to exist.

David: I guess my role in life is to look at problems and try and find solutions to them. And metabolisms has lots of problems; there’s lots of solutions that are still needing to be developed and part of it has come from the scale. I didn’t think it was going to be be such a large, unwieldy set of compounds that we had to work with. I didn’t realize it was going to be so complex. – And that metabolism isn’t just simply catabolism and anabolism. Metabolites have a role in signaling, metabolites have a role in health. They have a role in disease. They have triggers at so many different levels. Before I started we only knew about oncogenes. Now there are onco metabolites. So there are a variety of endogenous toxins, uremic toxins that seem to lie at the root of most chronic diseases. Things that we just didn’t know 20 years ago when we started in this journey. It’s always surprising; that’s what keeps it interesting for me exactly.

Alice: And this takes us to our last topic. I wanted to discuss with you the applications of metabolomics in the clinic or to clinical research. You are involved in projects that target the application to the clinics. Do you want to tell us first about the different projects you’re working on at the moment or things you can talk about? Are there diseases or specific types of applications where you really see promise for metabolomics or where you’re working to use it in the clinical setup?

David: We often underestimate the role of metabolomics already in the clinical setting. – We call it clinical chemistry and clinical chemists don’t want to call it metabolomics. My first encounter after discovering that they could use NMR or for mixtures was dealing with some inborn areas and metabolism and that’s when we first applied these and looked at urine samples of kids who had serious metabolic disorders. It was quite striking just to see the differences and how these compounds can be picked up. Of course, these days we don’t use NMR. It’s mostly LC-MS, but this is one of the great wins from the perspective of metabolomics. We have, I’m not sure, about a hundred thousand LC-MS tests done a week around the world for newborn screening. There are more people who’ve had and will have metabolomic tests than will ever have genetic tests. And it saves lives! It makes a profound difference. So metabolomics is already in the clinic. Metabolomics is already having a profound impact in people’s lives. At the beginning of life, it’s, playing a role. I think metabolomics can also play a role, in midlife or towards the end of life as we try and identify some of the things that scare us all: Things like cancer, Alzheimer’s or diabetes. And I think there’s a real role for identifying the first trends towards pre-diabetes, before Alzheimer’s there’s mild cognitive impairment before sort of full-blown cancer, there’s stage one cancers.

My focus over the last number of years is to look at identifying metabolic biomarkers that are predictive or that identify the earliest stages of disease. Because I think the role for metabolomics can be in prevention. Genes tell you what you might have metabolites tell you what you do have. Being able to monitor over a person’s life course and about changes allows you to pick up conditions before they manifest. I think this is the role that metabolomics really needs to play. Whether you call it in vitro diagnostics, predictive diagnostics, or preventative medicine – I think metabolomics can play a real role in a shift from reactive medicine to preventative medicine. I think it can play a real role in wellness and health as opposed to treating disease. And I’d like to see people use metabolomics different from a standard diagnostic test, “one chemical one disease” to more like hundreds of chemicals to assess your general health state. We are complicated organisms and our health is also made up of many components. And, I think metabolomics gives you that opportunity. Our focus is on preventive early stage, clinical markers. In all cases, it’s really important to have quantitative values, because without those you can’t translate them into reference values.

You can’t compare them to reference values, you can’t translate them into clinical diagnostic tests. You can’t use them in any kind of medical field. This is why we’ve always worked in quantitative targeted metabolomics. It’s made it very easy for discoveries we’ve made to translate into clinical applications. I think it’s been quite productive for us.

Alice: You mentioned, having, let’s say a hundred metabolites for one disease. So does that mean you would favor more patterns or like larger signatures as. Biomarker for maybe more for complex diseases than others.

David: I think from a diagnostic perspective you want a small number of biomarkers. It makes it more useful from a testing perspective but the concept with metabolomics is to go to more than one marker per disease. So you could have two or three or four, and that gives you greater specificity and sensitivity. But from a health monitoring perspective, I would view it as important to be able to look at several hundred metabolites at a time because you don’t know what someone’s is going to get. And so if you’re forcing someone to do a hundred blood tests to do a hundred different single marker tests you’ll bleed them dry so if you could with a drop of blood measure 500 compounds, which are reflective of most of the common disorders then you have a way of measuring health, not necessarily measuring a specific disease, but measuring health and that’s why that broad coverage is useful from the wellness perspective. Whereas from the disease diagnostic, you just want to measure a couple metabolites at a time.

Alice: And maybe also it helps, because if you have a single metabolite, it might be changing for different reasons as well in a given individual and this takes us also to this idea of the variability between different individuals.
That’s is something that sometimes people see as an issue with metabolomics. So that also for a given individual, if I eat something different I might have different levels of amino acids, which are used for many conditions. But also, this variability is something that can be very useful and that can give us a lot of information, and that is one of the reasons why metabolomics is a great tool for precision medicine, maybe more interesting than other types of omics. You’ve written about precision medicine in metabolomics quite a lot. What, what’s your position on this? Why is metabolomics interesting for precision medicine and what works for metabolomics and what maybe makes it more difficult?

David: You brought up the point about variability. There is a fair bit of human variability, and I think this is from the cross-sectional side. When we look at people just at one time point we see variability. Over a longitudinal level we don’t vary a whole lot from our set points. And so I think the central advantage of metabolomics is this ability to measure over time and to measure those changes over time and to have a reference point potentially even say “today I’m very healthy”. That’s my reference point. If you’re in your twenties, that’s probably when you’re healthiest – that’s a good reference point. As things drift up or down good or bad, that’s something you can track.

Your genome doesn’t change. Maybe you’re born with a bad genome, or a good genome, but it’s just hard to know when those things might suddenly lead to a problem or maybe it’s just a threat you live with for your entire life and it doesn’t happen. So I think from precision medicine perspective, the fact that the metabolome can be detected first days of life, that’s when we do newborn screening to the end of life allows us to see how things have evolved. I think it’s ability to predict conditions or prognosticate makes it unique. Proteomics can help, but the advantage that that metabolomics has over proteomics right now is that metabolomics is much more quantitative and the assays are much cheaper. They’re faster to perform. And when you look at the number of biomarkers that are used in the clinic, the overwhelming number are small molecules. The fact that they’re quantitatively measurable and fast reproducible gives an advantage. ELISA tests are reproducible within a lab but not across labs and very tremendously with matches of antibodies other conditions. So that metabolomics is the spawn of analytical chemistry. Analytical chemists are focused on, not only accurately identifying, but accurately quantifying things. If we remember where we came from as analytical chemistry, I think that it opens the door for much more useful clinical work and much more precise measurements for precision medicine.

Alice: Besides the new newborn screening, have you already seen from your work interesting new applications of metabolomics for diagnostic or for screening? Are there already things out there for those diseases where the work is still a bit difficult? Like, for example, to detect a disease a few years before it’s really too late? Do we already have good examples of the application of metabolomics?

David: We have applications that are published but not in use. And I think this is a real problem for the field of metabolomics and for the field of medicine. I think it could be regulatory problem but there’s also an inertia. It’s very expensive to go from a discovery or a publication to a validated biomarker that is used in the clinic. The funding systems that we have don’t really encourage people to do that. If, as a scientists, your worth is measured by your publications, not by the number of biomarkers or even the number of drugs you’ve developed.

Biomarker identification isn’t a highly profitable business. In some cases, if you identify a biomarker that is better and cheaper, is not accepted by a lot of people because the people who are making a lot of money from the more expensive biomarker are not happy. I think there’s a lot of constraints that undermine the ability to translate really useful biomarkers into practice. And some of them are structural and institutional and it’s a real shame. I don’t know if there’s things we can do to change that. But if there are funding agencies listening, it would be nice if they appreciated the importance or created mechanisms to help translate biomarkers because without biomarkers, precision medicine really isn’t possible.

Alice: Nope. Maybe we need a specialty task force for this. Yes. To push metabolomics forward. Then I only asked you to think about your favorite metabolite a few minutes ago. So have you found one and can you tell us why it would be your favorite metabolite of the day? Because I guess you have a lot.

David: Yeah, there’s 250,000 to choose from! One, that I found really fascinating, it surprises me every day, is the largest amino acid Tryptophan. And it was the only one I could remember the single letter for, because its last name. It’s one that has remarkable roles. It plays so many regulatory and signaling roles in the body. So it’s not just simply a proteogenic amino acid. It’s an amino acid that’s used in controlling the immune system. It’s an amino acid that plays a role in mood and neuronal signaling. It is an amino acid plays a role in formation of melanin and melatonin and pigmentation. It seems to play such a central role, but it is not only a good amino acid, it can also be turned into something bad. And the conversion of tryptophan into indole and indoxole sulfate which is a pretty serious toxin. It seems to have a role in anxiety and depression and chronic kidney disease and probably Alzheimer’s disease. The fact that you can have a single molecule with use and utility spanning so many things that’s both good and bad. That is fascinating to me and I think it embodies what I think we will find out about most metabolites – that they have all these roles and they’re not just simply fuel or they’re not just simply building blocks.

Alice: Absolutely. I completely agree. And probably the next revolution (it’s already started) but you started with expecting a few hundred metabolites and found out that there are much more, and probably now we think each metabolite has a few functions, but then we’re going to figure out that it’s all linked together and everything can do anything if it’s in the right context.

David: Exactly.

Alice: This is going to be fun. Especially for those doing the interpretation. Yeah. Thank you very much for taking part in this podcast. It was a pleasure to speak with you.

David: Thank you, Alice.

Robert Ahrends

Robert Ahrends @UniVie

Favorite lipid
lyso-sphingomyelin

Discussed paper by Peng et al.:
Identification of key lipids critical for platelet activation by comprehensive analysis of the platelet lipidome

Lipidomics pioneers mentioned in the interview
shorthand nomenclature developed by Gerhard Liebisch

With the corresponding text from the interview:
There was great work done on lipids in the 1980’s and 1990’s already. People like Britta Brüger, Andrej Shevchenko, and Wolf Dieter Lehmann were pioneers of lipidomics and translated nanospray methods for the analysis of lipids. In the early 2000’s, Britta Brüger published an application of nanospray for direct infusion instrument, which made it more interesting for membrane biologists such as Kai Simons or Felix Wieland who also works in Heidelberg. They then adapted this for their own research on lipid transporters. And since then, it became clearer and clearer that mass spectrometry is the tool to go with for lipid analysis: you can analyze a lot of things simultaneously and do it in a quantitative context.

International lipidomics society – ILS

Episode Transcript

Alice: Today on the podcast, I’m joined by Robert Ahrends.
Hello Robert, you studied biochemistry at the University of Giessen in Germany and got your PhD from the Humbold University in Berlin. After that, you went on to work for a year at Agilent, developing analytical methods, and then went on to Stanford for a postdoc in chemical and systems biology. You returned to Germany in 2013 to be a group leader at the University of Dortmund.

Since 2020 you are at the University of Vienna in Austria where you’re an Associate Professor at the faculty of chemistry. Your work combines mass spectrometry and data analysis to decipher the metabolome, especially in contexts such as cardiovascular disease and signal transaction but you do have other activities, like community building activities in the world of lipidomics.
What would you add to this by you?

Robert: So, was the Institute for Analytic Sciences (ISAS) in Dortmund, not the Technical University.
In general we are interested in lipids in a very holistic way.

We are interested in the lipid itself, but we are also interested in the metabolism around these lipids – So what enzymes are shaping membrane changes, signal transaction, … And for us it’s very important to always have a functional angle on this. What does it help us if we’re studying the lipidome or individual lipids if we do not find the root of the changes and put it in context.

Alice: Yes, and this is something we will, we will discuss today together. We will talk about one of your publications from 2018 about platelet activation and the role of the Lipidome in that context.

Before that, maybe more general notions about lipids, as we all know, if we work in biology, lipids have been poorly considered at the beginning of biochemistry. They’re one of the last classes of molecules to be investigated with omics, for example, What do you think that is? Even though we have a lot of lipids in our membranes and lipids are a huge part of our brain. So I you would expected people paid attention a bit earlier. Do you think there’s a specific reason why lipidomics took a bit longer?

Robert: No, I think there was great work done on lipids in the eighties and nineties. However at the time, mass spectrometry was not a key parameter of many molecular biology labs. With Britta Brügger, Andre Shevshenko, Wolf-Dieter Lehmann, Kim Ekroos who were pioneers of lipidomics and translated actually nano spray to the analysis of lipids.

There was a very interesting publication around the year 2000 from Britta Brügger who applied electrospray for direct infusion experiment and this then became more interesting for membrane biologists, such as Kai Siekmann and or Felix Wielan, which are also working in Heidelberg.

They adopted this for their own research on lipid transporters and things like this. And since then, it’s becoming more and more clear that mass spectrometry is the tool to go for lipid analysis – because you can analyze a lot of things simultaneously and you can put this in a quantitative context.

Compared to metabolomics and proteomics [in lipidomics] you have to do quantification because lipids are acting in bulk and not as single entities. For proteins it is easier to put in context because you have, for example, a nuclear localization sequence on the end of the protein and you know that this protein goes to the nucleus.

For lipids, you don’t have something similar. And lipids are most often sticking in the membrane together. So if you analyze changes without standard, you do not know what’s the impact of this certain lipid on the membrane fluidity, stiffness of the membrane, and other things. So reporting in concentration is key. And therefore, it took a little bit longer because at the beginning not much standards were available. Now it’s becoming easier because we have big companies producing the standards and we stepping forward.

Alice: You’re mentioning the importance the composition of the lipids, for example, for the fluidity of the membrane. This is something also where we needed this level of detail in the analytics to be able to know how long the chain lengths were and how many unsaturation we had in certain complex lipids to have a way to interpret also the effects that it has.

If you talk about membrane lipids, if you know that there’s an increase in the number of unsaturation or a change in the whole family of complex lipids with a certain type of fatty acid composition, then you can start interpreting the effects they would have on the qualities of the membrane itself, right?

Robert: The more information you can get, the better. There is a shorter nomenclature, which was developed by Gerhard Liebisch and colleagues where you go from lipid category level down to the isomer level of lipid – At the moment we are not at the isomeric resolution, we are somewhere in between where it’s possible to do MS2 experiments easily and get the fatty acyls with the number of double bonds and chain lengths.

This brings us already a big step forward because with this information you get an idea of what lipids are connected to the backbone. This information makes it easier to interpret membrane physics but also helps you to pin down if a certain lipid acts as a precursor for a signaling molecule. From there, you can also go deeper down to pinpoint the location of the double-bond. Though there are instruments which can deal with this, however, this is not broadly applied in lipidomics.

Alice: At the moment, there’s still a lot we can do with the methods we have. Your paper that we will discuss is a good example of this, I think. So you really start from the global lipidome and then progressively focus on the specific species of lipid that is going to be the beginning of a whole pathway. That we will discuss in a minute. Before we get into that detail, I wanted to ask you: What is a lipid for you?

Robert: As I started, there was a definition of a lipid as a hydrophobic biomolecule but this is a problem now because we are dealing with all the water soluble and water insoluble entities at once. If you thinking about a fatty acid, which has some hydroxyl group, this will be definitely water soluble but if you goes into a cholesterylester; this will be water insoluble.

You have to cross some 25 log orders of magnitude in water solubility. So I think people more and more say lipids are organic molecules to prevent to say they are hydrophobic entities. However, if you are thinking in the classical way about it I would say: Hydrophobic biomolecule, because even if it is water soluble, it will be less water soluble compared to molecules like phosphate or an ATP or something like this.

Alice: This is important to know for sample preparation. Doesn’t it. If you prepare to do analytical chemistry to study lipidomics more than small molecules, for example, do you use different methods to prepare your samples?

Robert: This is true. There are 1-phasic extraction methods, but most of the people in the field prefer biphasic extraction protocols. This has the benefit to first precipitate all the protein. On top of this, compared to the monophasic extraction protocols, you can separate more hydrophilic molecules from hydrophobic molecules. This doesn’t mean that you have to throw the hydrophilic phase away. You can still analyze it. But you separate out things to put them in different workflows. A proteomic workflow can go there.

The lipidomic workflows there and then a metabolomic workflow. On the other side, this metabolite fraction or the more hydrophilic fraction still contains a lot of lipids such as lysolipids, highly phosphorylated lipids, sugar lipids. Most likely will find them in this fraction. But you get any cleaner detection; there’s a reduced background noise. I would recommend that.

Alice: from the background I have, I remember we had different processes for different omics, and I think the most important thing is to be aware of which method you use and where it might make you lose certain analytes.

For example, we tried to combine proteomics and metabolomics from the same samples, and we did that, but you have to be aware that when you use a special solvent or a certain method, then you might lose a certain class of metabolites or a certain part of the sample to the extraction method you have. And doesn’t mean you can’t use it for anything else, but you have to be aware of the limitations of your methodology.

Robert: Exactly. And there was a nice publication Christina Kromann did from my lab in Molecular and Cellular Proteomics and you have to care of the partitioning in between the two phases. However, if you add a standard directly at the beginning, a lipid standard or a metabolite standard, it is accounted for. It helps you just have to detect the molecule and then it’s accounted for. This is a great deal when you are doing quantitative analysis. You just calculate it back.

Alice: You mentioned these membrane forming lipids and also signal transduction-related lipids. Do you want to give examples for those people who are not too familiar with lipid classes. What would be more on the signal transaction side, or that would be less part of this structural world and more of the effective world.

Robert: In general, if you think about how signaling can work, there are two different direction, right? So one is a signaling molecule binds to a receptor. And therefore, it should in most cases be water soluble. So we are talking about molecules such as oxylipins, phosphatidyl-1-phosphate and things like this. – On the other hand, there are signaling pathways, which called the unfolded protein response, where the membrane has changed.

And the change of the membrane or the distortion of the membrane is doing something with the dimerization of proteins and then downstream is a signaling pathway triggered which not directly involves a signaling lipid, however, it’s still a signaling pathway. So if you focus on the first one, there are all the different lysolipids there. It doesn’t matter where it’s coming from. It can be the sphingolipid metabolism or from the glycerophospholipid metabolism. There you most often find transmembrane receptor protein receptors. They are doing signaling.

Alice: And the lipids can be both, the ligands to the receptors and they can also be the effectors downstream of what happens once the receptor has been activated. For example, in terms of ligands, I’m working at the moment a lot with bile acids and if you had asked someone 30 years ago what bile acids do, they would’ve just told you: They just help us to absorb fat.

Then, in the late nineties, it was discovered that there were ligands for certain nuclear receptors and that revolutionized the world of the research on bile acids and suddenly they were interesting again. So we might discover other things like this for other lipids as well. We still discover relatively recently things that we thought we knew and then find entirely new ways that they can regulate biology as well.

Robert: Maybe I forgot something. We also have also Phosphoinositols. They’re activated by a pathway and they’re shut down then becoming diacylglycerols or other metabolite entities, and then act again as signaling molecules on calcium signaling. This is somewhere in between membrane lipid and a lipid is binding to receptor.

Alice: There are lots of possibilities. So before we go to the specific paper we’ll discuss, I would like to ask you a generic question about how you handle data interpretation in general. Let’s say you have a typical lipidomic data set. Do you have a strict workflow of how you analyze the data or do you tailor this to every paper and every study and question that you want to address?

Robert: This is a complicated question. We are tailoring this most often to the model system because somehow there is always a bit of knowledge known. For example, if you’re working on synaptic junctions with the synaptic cleft, you have to consider gangliosides.

So if you are in the platelet regime you have to consider oxylipins and molecules like lysosphingolipids, which are known to do something to platelets. So it has to be a little bit tailored and there we try the following thing. So we have a direct infusion essay where we cover all the phosphor lipid triacylglycerols and cholesterolesters (the bulk lipids), and then we have special targeted workflow for sphingolipids, oxylipins, PIPs, for example.

Alice: These targeted workflows, they do special kind of combinations of lipids because they have specific structures and combining them differently because of the way they are made or what kind of specific workflows are you talking about there?

Robert: So the workflows are more or less analytical tailored. They’re about accessibility – The extraction methods are the same. But for sphingolipids you would like to get rid of the phospholipids. Okay. So you keep all the esterbonds with alkaline hydrolysis.

For PIPs you would like to extract very sour. So you reduce the pH and then things like this. So they’re a little bit different requirements on the detection. You have bulk lipids which are very abundant and then other lipids. PIPs are very low abundant and certain sphingolipids are also very low abundant.

For them you have to go into a more a targeted direction whereas, if you go for general lipidomics you can stay more untargeted, discovery like extraction by covering with standard certain lipid classes. – You always want to work quantitatively.

Alice: Do you often then go to the route where you would start with untargeted to figure out which are the most relevant classes of lipids and then maybe go more towards targeted.

Robert: Somehow we are we doing this? The bulk lipid analysis is untargeted, anyway. For the targeted things, you cannot really do untargeted because you are losing sensitivity. What we do? We open the panel using lipid creators, a software tool, which we have developed.

Then we are putting a lot of information inside and do a screen. This is not really untargeted. However you can cover in some runs a lot of different lipids with the speed of modern instruments. And then we shrink it down to a defined target list on the end of the day of lipids, which are in the systems. The key is that we can do the permutations of the target list at the beginning to figure out what is in our sample.

Alice: This is to analyze the sample and then identify the lipids. Then let’s say you have now a matrix of your data where you have all your samples or your piece that you’ve identified. What do you do from there? Do you do more data driven analysis where you do a lot of correlations between different sets of lipids or do you also work on a more knowledge base where you would use pathway analysis or chemistry driven analysis where you group things per class? Do you have a typical workflow for that?

Robert: We can create this matrix because we are quantitative. So we can get all our workflows together and we creating a big matrix. Currently, we do two things. The first is more a mathematical correlation driven approach (using Pearson correlation get different clusters and different clusters using correlation and anticorrelation then gives you a network).

So it’s still a network and you can easily visualize what lipid classes are there and what is changing there? With one look you can see the changes in the system. You don´t fully rely on this information but you also don’t lose information. This is the first approach. The other approach we call the chemical space. There’s a model developed by Dominic Schwudtke and here you can calculate distances in between different lipidomes. This helps also, if you go across species which have different lipid entities, so different lipid species in inside. There are species of animals and species of lipids.

Alice: I also feel like I have to explain which species I’m talking about very often.

Robert: Since you have this chemical space model you can start comparing the things and visualize the things. At the beginning we started, of course, with some reaction pathways and network biology-driven things. However, we came very quickly to an end because with lipids, not all the information are in reactome or in KEGGs. There are some things already are now getting more and more established.

However, there’s still a lack of knowledge because nobody doing the hard groundwork where in molecular biology to find out what a substrate specificity a certain ceramide synthesis has and what not. This knowledge is missing in biology and therefore the databases are not becoming better (The information is just not generated). This is a problem. So therefore we decided to go to more data driven approaches.

Alice: And this is something that you can afford as well when you have a relatively large data set where, I think, in the paper we’re talking about 400 lipids or something that were measured. So you have a matrix that’s big enough that something comes out of it. Of course, when you measure 12 metabolites, unless they’re very strongly related to each other because you chose them like this you have less chances to find something interesting. With larger data sets it starts to really be powerful to use this kind of correlation.

Robert: But even if it is a larger dataset. So the 11 lipids, which are highly correlated, which was a pop up in this data, said as a very dense of course goal. Of course. Yeah. And then you’ll see it straight.

Alice: Let’s talk about the paper. The paper I wanted to discuss with you is titled: “Identification of key lipids critical for platelet activation by comprehensive analysis of the platelet lipidome” you are last and corresponding author on that paper with Oliver Borscht. And this was published in blood in 2018.

I wanted to talk about this paper because there are a couple of figures I really like about it. It’s a good example of a common approach that is to take a large data set and that I’m really mostly talking about the lipidomic data set. So there’s a whole story behind it, about the biology that I will let you tell in a second.

The thing I really liked about the data approach in here is that you have a kind of

1. Overall description of the omic data set
2. Very tailored tools that allow us to look at it from different angles to have an idea of what are the important, features in it. What is the spread of the data set in the chemical space and all this kind of things
3. Another couple of figures where you compare with and without activation of the platelets to see what is the impact on the lipidome and
4. Look at one specific pathway and see by looking at the different species of the different lipids involved in the pathway, how you can actually follow the pathway when you activate the platelets.

I found it really elegant. Could you say a word about the biological background about this study? Just what were you trying to look at? And, why was lipidomics an interesting method to look at it in that system?

Robert: First, thanks that you like to paper – Nice to hear. Why we did this: Because lot of things happen in the platelet lipidome! Platelets are changing drastically the morphology upon activation. Fatty acid pathways, the PIP pathway, phosphatidylserins which switch to the outside, different phospholipases involved in calcium signaling linking back to oxylipins, autocrine feedback loops to the entire system where lipids are responsible for. So really a lot goes on.

However, dynamic changes also mean instability. So, you don’t want to have something dynamically changing where you have to keep the sinks intact, at least for a certain time. This was contradictory in our opinion and therefore we went in to make is this Quantitative platelet Atlas. We used different stimuli to look how big are the changes? What is the impact on the membrane? These were the interesting things to find out.

First of all, which percentage of lipids are changing either within five minutes. In five minutes you have clotting. (You poke yourself. How long are you bleeding? Five minutes. – Not longer than five minutes. Depends on the size of the wound. But you should not bleed very long. So it must be a quick process. Five minutes seems to be okay for it. We discovered that just 20% are changing.

And this also stimulus specific. There was another very interesting catch that just 20% lipids making up 80% of the entire lipidome. And a lot of these lipids were containing PUFA lipids. Especially here arachidonic acid. This was then the next step.

We remembered all the classes at university in biochemistry – How signaling of oxylipins is working. You need a precursor – And this precursor is arachidonic acid. So we focused subsequently on our arachidonic acid signaling (thromboxane, prostaglandins, …).

Alice: Yes. Let’s look first at the description of the dataset in general. Kind of this landscape view of what the dataset is; what does it look like? And this in the paper is figure three. So there are a couple of graphs and the one you mentioned, for example, where you see that these 15 lipids that are actually corresponding to about 70% of the lipidome in the platelets and the rest (the other 380 something that you measured).

So it’s really intense. So cholesterol being the first component. And then, as you mentioned, a lot of PUFAs. You also compare human and mice platelet lipidome and check to see if you find the same kind of species in both. – Sometimes it’s the case. Sometimes there are differences. This is also interesting when you think, okay. We might be interested in studying, plate activation in the mouse model. So we should be aware that there are things we will see there that we might not see in humans and vice versa.

This is always the main thing you have to keep in mind when you use animal models to try and better understand human biology. And then there’s this beautiful figure 3e where you break down the lipidome into different classes and beautifully visualize with the different classes and the order of magnitude of concentration that you find for each member of those subgroups of the lipidome.

That really gives a flavor of the spread of concentration that you already mentioned before and also of the variety of different types of lipids that we can look at. – Because, if like me, you know little about lipids before getting into this field; just what you learned in school and on the benches of the university – this is not at all what I remembered from my classes 20 years ago when I learnt biochemistry and molecular biology. It didn’t go into that detail. And I hope that now we’re including this into the teaching.

It’s really a beautiful new world of biology that is super interesting to investigate. In omics papers you often have this one or two figures that give us the vision of how broad the data set is, where it can take us, and what makes the main components of it. – So you also see the contribution of the different classes of lipids to the overall lipidome. The same thing is in the next figure when you compare then activated versus resting platelet lipidome that you see then which are the classes that change and in which direction does it move.

And one figure I really liked was figure 4c where you make this very simple graph where you show the lipid-lipid correlation within class and between class. This is very interesting thing to see because you have correlation between one lipid and another lipid, of course, but if that lipid is from the same class, then that would make sense (e.g. you have a whole pathway or a whole metabolic biosynthetic pathway that moves in one direction), but when it’s between classes, this can be for different reasons.

So when you have a correlation between classes it could be a typical example if they share a common fatty acid that they incorporate in different types of complex lipids and then this kind of makes them move together because we have an excess or lack of a certain component.

Robert: Exactly. That’s something coming from the outside.

Alice: It’s like one of the small building blocks is coming to all those groups and they all move in the same direction, which, you can guess, or at least you can think of looking at it when you do this kind of simple correlation matrices, looking at the, at what you can see within class and between class. I thought this was a very simple graph to make but it can be really informative.

And this is something that’s really specific to lipids. If you work in small molecules, you can make a similar kind of, graph but you wouldn’t come to the same conclusions because they usually don’t share the same building blocks. What I’ve also found interesting from at biological point of view is that when you activate the platelets with thrombin, CRP, or both, you don’t get the same impact on the lipidome. So platelet activation is not just platelet activation.

Robert: That´s true. There are different receptors; a lot of them.

Alice: It’s like it’s platelet activation, but it’s not exactly the same. And actually, in the figure where you focus more on the oxylipins and the prostaglandins, then you see that it’s only when you combine both that you actually get this really intensive impact on the arachidonic pathway.

To go back to figure 4 where you compare activated versus non-activated platelets you have one of the most efficient ways of analyzing large data sets is this correlation matrices and then finding clusters of lipids that correlate together and looking inside those clusters. And here I always focus a lot on the visualization because I liked the way you visualize this.

So you have the typical correlation matrix and then you extract each cluster and open it to see what’s inside on the class level in the next graph. And then you take these components of these clusters and put them inside cytoscape to build the networks and then see how the networks light up with the different activation strategies that you have.
My main question is it how you always do things or do you do a lot of trial and errors and a lot of exploratory analysis and trying visualizations to find most efficient ways to show it in the best way in the figures.

How does that work on your side?

Robert: This depends on the data that we are looking at. However, if you go through the data, there are certain key features which just jump into your eyes. Then it’s a long way of proving this and connecting this to the greater lipid environment.

And I think, such tool such as Pearson correlation cluster approach are just standard tools in omics but together with these networks it is quite powerful to visualize lipidomes. We don’t always do it in such a way – This is just data driven.

Say we would go, for example, to a deeper level and the human plate lipidome and mouse lipidome would be completely separated. You cannot do this anymore because the lipids are not shared. So if you cannot compare A against B you must come up with something that adds up. Therefore, for example, you could use the chemical space to identify chemical features of a lipids.

Even if the chain lengths are different and still you somehow get the same hydrophobicity, the chemical space model would better reflect what’s going on in the system. In this example, a different class of lipids is taking over the function of a given class of lipids in the comparing system.

However, if it’s totally the same set, then this is the perfect to go. If they are more distinct, I would slightly change the direction. At the end of the day, of course, you have to adapt this in the best way to bring out what you want to prove with the experiment.

Alice: Once you have the result of the correlation and the components of the cluster, you could also make a heat map for example, and you could show how the heat map lights up with the different activation methods. – Would that be the same for you or do you see a benefit of using the network? There is an inherent component of the network that shows the hotspots and that shows, which ones are heavily linked to others.

Robert: If you would use hierarchical clustering you would consider everything. For certain components you would’ve a nice cluster on a certain point, as you can see it already in the Pearson correlation and Cosine clustering on top. Then you have this cluster there and still you ask yourself “What is going on there?”, “How good are they correlated?”. And you don’t have the cutoff – With the correlation, you can define a cutoff, just go inside and look for the highly correlated or anticorrelated molecules at once.

You look at them at once. With the other visualization you cannot look to the things at once; they’re on different spots. And here you can do this at once and on top of this, you can put mappings of classes onto this network before that lets the reader instantly understand what classes are involved.. In your hierarchical clustering map you can do this, too, with extensive labeling but I think this is already information overkill for a visualization. To summarize – On the one hand we reduce information in order to get the right information to the reader at the right time.

Alice: It’s a communication game in the end. How long does it take to make an analysis like this? In terms of scale what would you say in general? I think people underestimate how much time it takes.

Robert: The most challenging thing is to get the matrix done. If you have the matrix right, then it’s fast. Then you talk about two weeks to get to something which you can present. Just to get some data to look it is just some hours because it’s all scripted. In order that you are happy with this, I would say, two weeks. – But the data analysis upfront is rather two months.

Alice: I expected something in the range of months like overall after the experiment is done and everything like it does take a lot of time to process data. Of course, it always depends on the question you want to answer in the first place, but still, even if you know what your question you want to answer there are still different ways to get to the answer.

Assistant professor at the Institute for Computational Biomedicine, Weill Cornell Medicine, New York
Krumsiek lab @Cornell

Favorite metabolite
2-Hydroxyglutarate

Discussed paper by Krumsiek et. al.
Gender-specific pathway differences in the human serum metabolome

Getting started with Gaussian graphical models (GGMs)
MoDentify R package

Modular metabolomics pipeline
maplet R toolbox

Open access multi-omic tools discussed (from other groups)
Multi-omics factor analysis (MOFA)
OmicsAnalyst

Other resources
Krumsiek lab github repository

Sign-up for The Metabolomist e-mail list

Episode Transcript

Alice: In today’s episode of the podcast, I am joined by Jan Krumsiek. Jan, I start by introducing you, because some members of the audience might not know you, but I’m sure some of them are familiar with your work.
You studied bioinformatics at the technical university of Munich in Germany, and then you did a PhD at the Helmholtz Center in Munich as well. You stayed at Helmholtz for a few more years as a team leader and junior group leader in Systems medicine of diabetes. In 2018 you joined the Weill Cornell medicine in New York, where you are now an assistant professor. Your group focuses their work on the development of new bioinformatic tools to analyze metabolomics and other omics – so to integrate datasets. Would you like to tell us maybe a bit more about the work of your group at Cornell?

Jan: Yeah. So when I came here, I think it was still that metabolomics wasn’t used as much, or people were using it as a tool, for metabolic research, but they didn’t have necessarily the methods. Together with a colleague from Qatar, Karsten Suhre we convinced the department here that they need a computational metabolomics person. That’s sort of the reason I came here and, as you said, we are developing methods all the way from pre-processing data, quality control, all the way down to pathways and networks. The only thing we don’t do is working on peaks and spectra directly from the mass spec[trometer] – usually we rely on the person on the platform for that and take it up right after that.

Alice: And which platforms do you use to develop those, those tools?

Jan: In the very beginning, a lot of our work was based on MATLAB. There was not even a specific choice – It was because my PI back then came come from the second processing physics field and he just was a MATLAB person so we were all doing that. If I could go back in time, I don’t know if I would do it the same way simply because MATLAB is very expensive, and the toolboxes are not necessarily made for metabolomics analysis. So these days I would say we do 95% R!, maybe 4% Python, and the rest, sometimes whatever we need in a specific scenario. The majority of the work is done R! based because it’s free and it is supported by the community a lot.

Alice: You worked on diabetes and also on cancer and also other diseases. How do you choose which disease you work on? Is it based on the collaborations you have or is it also your own, driver driving force for certain, diseases?

Jan: A little bit of both, but I would say. A lot of times it is driven by the opportunity of collaboration. We have to admit that. We worked in diabetes and then cancer and also Alzheimer’s disease. So what else? It is almost all of the diseases – sounds a bit much, but they do come together at the level of metabolism where metabolomics methods can be used for 80% of the same things. Right? But then the biomedical applications is driven by who’s there to work with you. Here at Weill Cornell in New York, we have a lot of oncologists and oncology researchers and clinicians, and that matters. If you do have people that are specialized in the field around you and if you can convince them to obtain patient samples for you and are ready to run a project with you.

Alice: So let’s move to the first topic I would like to discuss with you. It’s based on one of your papers from 2015, the paper is entitled “Gender specific pathway differences in the human serum metabolome” and I wanted to discuss this paper with you for two reasons. One is the methods: You use a combination of pathway enrichment and Gaussian graphical models or GGMs; and the second reason is the very topic of the paper, which is sex differences in biology, and specifically in metabolomics. In this paper you use two methods that I would call like one biology driven and one more data driven. This is how I would categorize them. From what I gathered from the methods, there is a pathway enrichment method, which is a kind of home-brewed method from what I understood. So did you design your own pathways in which metabolites belong and then do the calculations?

Jan: First of all, I like that biology driven and data driven. We always call it knowledge driven and data driven, but it’s the same concept of what you were referring to, so for the pathway methods the assignments of pathways of the metabolites are actually from the metabolon platform as they deliver them to us which is a very laborsome process. The have teams in the background making those choices that you later on work with.
And, there’s a lot of questions as we know that. So, you could do it differently. [For example] Why a certain glycolysis metabolite is in the carbohydrate pathway and not the energy pathway and so on. Those choices seem sometimes arbitrary. And you also get criticism from reviewers for that, but at least it’s something we have – That’s always our counter argument. At least we can work with it. Then, we can still ask questions later.

Alice: I think as long as you’re open about what’s in each list, then people can also make up their own mind about it. If you look at the map of the metabolome, you see that everything is connected to everything. So where do you draw the line; where does one pathway stop and the next one begins? Sometimes it’s not so clear.

Jan: Exactly. We have also worked with some versions partially from them partially from other databases where you don’t have that constraint of you have to annotate each metabolite with a single pathway, but you can do many – but life doesn’t always get better with that. So it’s really complicated if you have all these assignments, that you work on. So, the method that you were referring to, let´s say we have all the annotations and we can believe which metabolite comes from which pathway – What we found is that the classical pathway enrichment analysis, as we know it from genes, transcripts and so on, it doesn’t work that well in metabolomics, in my opinion. It’s being used and it’s this classical idea of enrichment. So, is there a pathway that has more hits in it? We’ve also had a PhD student work on that topic. It doesn’t work that well if your background is small. So we only measure a couple of hundred metabolites, maybe a thousands, and it’s not the entire genome as in the transcriptomic case and it creates some statistical artifacts. That will be bit too much to talk about here, but those [artifacts] matter, it makes a difference. So we come up with all these problems that these enrichment methods had in how we observed when we did it in this case – the male versus female analysis. So we came up with this new way, which was to aggregate them, to create a pathway score. That in itself is not new. Just the way we did it was new. So the idea is instead of asking, let´s say, “Does TCA cycle have an increased number of hits”, we ask “is the average concentration of the TCA’s cycle higher”. Slightly different question, but it matters. And that gave us these results that are described in the paper.

Alice: Okay. And then you had your other approach. That is GGMs. Could you explain for a broad audience, what GGMs are and then what that brings compared to the traditional pathway analysis or so other types of data driven analysis ways that’s especially interesting in metabolomics?

Jan: Yes. GGMs are first and foremost a statistical tool that don´t know anything about biology or biochemistry, but work with data. So what you can put into a GGM could be anything. The idea behind GGMs is correlation based analysis. It’s also been referred to as “guilt by association”. We do that all the time. So, if two things correlate (they go up and down at the same time) across samples, they must have something to do with each other. – That’s the idea of any co-expression correlation based analysis. The thing that GGMs add to that is that they get rid of confounding effects, that try to tease out using statistics what are the direct correlations and not what is the correlation partner of another correlation partner of another correlation partner distant in the pathway using regression based methods.
That method is not new – We didn´t invent that. And that has been out there for a long time and there are books about it. What was new was that we applied to metabolomics data. That’s easy. It’s a click of a button. What our research of the 10 years after the first paper (and including the first paper) was, is to prove that these correlation based networks that come out are not just pretty to look at, but that these statistical (not biological) models actually reflect pathways. Across platforms across tissues, across species we could show that and that was the big contribution of GGMs.
And you asked them about what is the advantage over regular pathway methods? So the interesting thing is that for GGM, you only need the data! For example, one of the biggest problems in Metabolomics is, as we all know, are unidentified peaks (you have a peak that you see consistently but don’t know what it is). You call it something X or different naming versions. You can do pathway analysis or anything with it. GGM will put that just in there. It’s just in the network and you can use it still. You don’t know what it is, but at least to see it in context and ask questions later, what did my.

Alice: And so one of my questions would be: How is it to interpret the results? If there’s a biological reason that could explain why they ended up in the same network.

Jan: Yes. We also use, and that was a paper we had later if that’s what you’re referring to, the information of networks actually to predict what they [the metabolites] are. It turns out to be so precise that you can read them from the context. With some inaccuracies it is reducing the number of candidates not telling you exactly what it is, but, it´s precise enough to show you what the metabolites are. And we found that particularly interesting, not just from a metabolomics centric view (We want to identify metabolites that no one knows about), but also because this is in blood, right? It’s not in a cell or in the liver or in the biopsy, it’s in blood and still the pathway footprints that we can find in it with statistics are still there so that we can recover all these pathways.

Alice: From your explanation of how the work happens under the hood, I was wondering how you take confounding effects into account. Does that mean that if you don’t specify that, the two sexes are two different groups, this would be kind of blended into the mixer?

Jan: That’s a good question. So what we’ve found over time, and that is more of a summary of published work and some unpublished work and work by other authors is that these networks that we reconstruct from the data are pretty stable across conditions in humans, for example.
So now you might say, can’t, we make a network for males and for females, two of them and compare, right. Or could we make, get diabetes into non-diabetes network, our cancer network. And at least from what we’ve found so far, it turns out that the networks themselves are remarkably stable, which at first was disappointing because we wanted to find the differential networks.

But it’s also actually interesting because that means I don’t have to really worry about it. I can use the same network as we used them, the paper for men and women alike. We just put it all together. And these confounding factors are more relating to other metabolites – because metabolites tend to all go up and down at the same time with each other. That creates a lot of correlation. That’s real, but it’s not direct. Whereas across different genders or age groups or disease groups, it seems as though those structures actually not confounded as much and are very stable, which is interesting, and it kind of makes sense. Metabolism is hardwired and then you just do make changes.

Alice: Okay. As any tool I expect GGMs have limitations. So what would you say are the primary limitations to the application of GGM to metabolomics?

Jan: Yeah, absolutely. There are two big limitations. The first one is sample size and we get that question a lot. You cannot do this in 20 samples. How many exactly do you need is a very difficult question because it also depends on how many metabolites do you have, how correlated are those? We’ve seen decent results with fifty to a hundred samples. Below that, I don’t know if I would use that method. That’s also why it’s often used more in human studies, rather than for example, in mouse studies, because you have very specific groups.
The other limitation is the first G in GGM (-Gaussian). So it requires normal distributions. That sounds very statistical, but it matters. Especially when you have clearly non continuous data such as a binary variable or an ordinal variable. For example, you cannot statistically put gender into the network as a node. That would be interesting to see gender float around with metabolites in the network and see where it attaches and what it does (you cannot do that you need a completely different type of methods – MGM (for Mixed distributes and graphical models), which is much more complicated from a statistics computation point of view.
So that is a limitation that is real and needs to be taken care of if you have mixed data.

Alice: Okay. For people who are interested in GGMs, do they have to call you to try and collaborate with you? Or can they play around by themselves already? Are there other software’s or code available for people to try this on their own.

Jan: There’s a couple of packages out there. If you just google partial correlation, there’s one called ppcor. There is another famous paper and package called GeneNet from the lab of Korbinian Strimmer, a German statistician which is widely used, and those two boxes are really one click (both packages are linked in the shownotes). Again, you have to take care of all those things, the Gaussian distribution that’s on you to figure out that that’s all okay. But then executing the calculation and getting the network from it is something that an undergraduate R! coder could easily do. No problem.

Alice: Okay. I also have the question what pushed you to study this topic in the first place? Is it looking at data and the experience with data that’s made you say, okay, there is something here. Someone should really describe these differences in the metabolome or how would it come from?

Jan: I wish that was the answer. What happened is, and I think that is an interesting anecdote: I was one of the first PhD students working with metabolomics data at Helmholtz. Back in the days we had biocrates data on the KORA study, and I was just trying stuff. I didn’t really know what I was looking for. And then my boss said: “Why don’t you try this thing? – The, what is it called? GGMs? Just, just try it out and see what happens.”

That is actually how it started. And I just pressed that button and looked at the Excel sheet, and then we inspected it for a while and was like, wait a minute. The first hits, those are all known pathway reactions. That seems like there’s something here. So I have to admit that this story that we tell (that you can use the GGMs to reconstruct the pathways from the data) is not what the hypothesis was for which we picked a method, but it was really the other way around. And that’s how it happened. I remember that Excel sheet actually at the very beginning of my PhD.

Alice: And about the relevance of sex differences in metabolomics. Do you implement that in your work now?

Jan: The interesting thing about sex difference is that it’s the most simple variable in your data – The most conceivable, one of the easiest to assess, to keep track of. So you always have it, with very little errors, usually. We can even, for example, in those thousands of samples, genetically verified that they crossed the right box on the questionnaire, and we have maybe one error in 1800 – so very easy to keep track off!
But as one of the biggest confounders of all, it really makes a difference. Men and women have very different metabolism. So the idea why we tackled that topic is that if you want to understand something complicated, like cancer, not saying that gender isn’t complicated, just the variability isn’t complicated, but if you want to understand cancer outcomes over time or diabetes complications or stages of Alzheimer’s.

Alice: You have a very good example recently with Alzheimer’s that in the paper from 2010.

Jan: Yes, exactly. And then that is how there’s dimorphism in the associations that the paper you’re referring to between metabolome and Alzheimer’s parameters. If you want to go to that level, you have to first understand what the baseline differences between the two sexes are. The other motivation, of course, there’s always a pharmaceutical idea behind it, in that metabolites do probe the metabolization products of drugs, as well. They can show you how well, how good you are at metabolizing something. And we know that a lot of medications these days still are dosed for adults, youths, and children – Not like men should take 800 mg, and women should take 600 mg, for example. And also for that some baseline research on how general metabolism adjusts in the elderly population in this case. Mostly healthy population.

Alice: I liked this a lot in the in the [Alzheimer´s] paper by Matthias Arnold: There’s a very elegant demonstration of the power of stratifying, the data based on sex and also based on ApoE status for Alzheimer’s patients where you see there’s one figure where you just compare Alzheimer’s versus control. No discriminations made whatsoever. And there’s no difference (proline was the metabolite that’s looked at, very basic amino acid). And I mean, I said, you think, okay, it’s nothing special. And then you start stratifying and you see, male and female looks a bit different. And then when you combine the ApoE phenotype and genotype and the sex, then you see suddenly this actually becomes relevant for women who have this specific genotype. And this was a beautiful example where sex is the point where the differences made.
In this case to address the metabolome of the female population with this specific genotype, because you’re grouping everyone together. And the same way, if you have a response that is so strong, that it might come from a small part of the population, then you’re going to generalize to everyone – And the majority is going to do something that’s useless for them, because the response is so strong in a small part of the population. So this, this was a beautiful example from the paper. I liked it a lot.

Jan: I’m glad you liked it. One of the major challenges in the field and the true challenges of the entire field are these stratifications. So in this case, the factors by which we analyzed, it were clear. It was gender was a good candidate to work on. And ApoE genotype known type was a major factor in Alzheimer’s disease. The factors by which the data were stratified were sort of obvious, but what if we are talking about something else? Something that you’ve never thought of like some age group or a very specific genotype group that we’ve never thought of? That becomes a real statistical problem, we can´t test all of those combinations of stratifications. I personally believe (though many people are working on those methods, of course), that there’s a lot of those hidden and cryptic associations out there that we just simply don’t know about them. And I don’t know what to do about that. Maybe we need really big data sets like the biobanks.

Alice: Is there something else you would like to point out about the sex differences paper?

Jan: There was one interesting story in the paper that shows how complicated it can be to analyze this type of data. We found differences between men and women of piperidine, which is a component of black pepper spice. Right. And it’s higher in men. And as always with correlation and causation, it’s easy to find, but not easy to explain. It could be, for example, that men eat more pepper, that´s conceivable. But it could also be that the metabolization is different because there are a known differences in cytochrome C metabolization xenobiotics that could also be, and we don’t know the final answer. Even in the paper we had to say, well, it could be this, or it could be that. And maybe no one cares about black pepper that much – But if this were a drug is really matters, right. This really matters because what we, what was the origin of those differences and it’s really hard to tell from observational studies. That’s still a major challenge for sure.

Alice: I did my PhD and my academic work in the world of toxicology. For me it would always be interesting cause it could be anything you’re exposed to. Just from what the chemical is, it even could be aftershave or something that has this pepper scent and then you put that on more when you’re a man than when you’re a woman and it’s present possibly with other chemicals that you expose yourself to without really thinking about it. It’s a really interesting field.
I know you’ve worked a lot with, the combination of metabolomics with GWAS as well, but you probably work with other types of omics – which type of omics datasets do you work with? (other than metabolomics?)
Jan: Over time we have worked with all of them, if that makes sense. The standards like transcriptomics proteomics, genomics, metabolomics, and then also epigenomics for sure. Very important topic. And some that are a little more specialized, such as glycomics. (That was again technology driven collaboration partner Gordon Lauc in Croatia) and some other more specialized aspects. The big set of the central dogma of biological omics, I would say all of them in some capacity.

Alice: I’m just thinking about this now on the fly: What is your view of epigenomics? How do you use it?

Jan: What I found most difficult working with epigenomics and I’m assuming everyone who’s done epigenomic research will have encountered this. While the marks on the DNA themselves are supposedly binary, you don’t have the local inheritance structure like with snips.
In a snip I can be somewhat certain that my neighboring snip is very highly correlated to me, which is the equilibrium that doesn’t necessarily count, at all, in epigenetics because it’s a chemical modification. So one mark could mean a lot and the mark three base pairs down could mean nothing. That really makes it complicated. We had one study in the context of type one diabetes and HLA methylation and the functional interpretation, summarization, aggregation of results across many marks on the DNA. I found much harder in epigenetics compared to snips. (which are also not easy, but I thought that [epigenetics] were way more complicated)

Alice: Then going back to metabolomics, you integrated with other omics. Do you have like an opinion of who plays best with metabolomics or is it the same for you? And you’re happy to combine it with anything?

Jan: The first thing we have to say, or we have to sort of explore is that integration is a word that we use and it could mean a lot of things. Right. While, for example, a collaboration partner approaches us with a pure metabolomic study, two groups, we do have a one size fits all approach for that. We have our standard pipeline analysis before the pre-processing and so on, and we spit out the pathways and then they can work that. In multi-omics I’m also being asked a lot. What is your standard approach? How do you integrate. This metabolomics data with the transcriptomics data that I’ve measured and it turns out there is no standard, one size fits all solution because there is no standard, one size fits all question.
What are you asking? Do the metabolites correlate with the transcripts? Okay. So we now can design an analysis for that. It could also be something way more complicated like “Are these enzymes regulating that metabolic pathway or does glycosylation of a protein make a difference in metabolites” for which, by the way, we don’t have any evidence. I think it comes down to that you need to know at least somewhat the question. Let’s stick with the correlation part, I think that’s the most intuitive do. The question is just “Do they go together” – Yes or no? – And then maybe also as pathways. I think the Omics technology that fits the best to metabolomics is proteomics. And the reason being maybe that it’s exactly the next partner in the cascade, transcriptomics is one step away but also the reason being that we often measure it in blood and blood proteomics, as blood metabolomics is something whole body. Come from everywhere, but transcriptomics something very different. Blood transcriptomics is immune cell white blood cell, mostly transcriptomics. So you profile the very specific compartment and that’s important. So when you say I do blood, let’s say transcriptomics and metabolomics, you picture them as just two steps away, you know, transcriptomics, proteomics, metabolomics next to each other, but they’re really not. They actually [spatially] compartments away. It’s more like liver-influenced metabolites or immune cell transplants.

Alice: In tissues, the picture might be different.

Jan: Yes. In tissue, the correlation is generally higher. We have an unpublished cancer dataset with our colleagues from Memorial Sloan Kettering cancer center. There we are exploring on the metabolome-transcriptome correlations in cancer tissue. Even there, it’s not as simple as you might think. It’s not that always the enzyme goes with the substrate and the product of the metabolite, as you would picture it.

Alice: This often comes as a surprise to people who have never done this work before. This is interesting to see then I guess you have to do a lot of explaining when you, work with people who are new to this field.

Jan: Yes. And to ourselves, too! Sometimes you wonder how that enzyme shows a lot of variation? It looks like there is something, but it does not go with substrate or a [reaction] product. And I think that just goes to show that dynamic regulation of biological systems is more complicated than the arrows we draw on a piece of paper.

Alice: So in terms of tools now, are there any software tools or programs with the code available? That you would recommend for people who are interested in integrating metabolomics with other omics.

Jan: I would always encourage people to go out and check the most recent reviews because as we record this today, tomorrow there will be five new methods. So there’s a couple of really interesting methods: for example, the MoFA method from the Oli Stegler lab that is going more in the direction of what I was referring to earlier (the cross correlation analysis). And then there’s several methods out there that do attempt to do joint pathway enrichment analysis. The Metaboanalyst platform famously now integrating multi-omics datasets, as well. But again, it really depends on the actual question at hand to then go out there and pick the right method.

Alice: From what I remember, the most complicated thing for us was to find the common language between the different data sets. Is translation the first step and maybe the most time consuming step in some sense?

Jan: Absolutely. So if you get a new dataset you can’t just go ahead unless you know that someone has already worked on the data so much that you can ignore pre-processing. I just take the data and work with it.
But even then, you find some result or something you need to understand better. Maybe a protein is not really what it’s supposed to say; you have to think about how the platform measures it; you got to learn every platform, every method, all the problems of it. – And we’ve, had this discussion internally in the lab that more often than not, we hope we can use the data blindly but almost every single time we have to walk back to the platform, talk to them. What’s going on here? What does this mean? Is this maybe a miss annotation? I don’t understand this. Why do I have all these zeros in my dataset?
Our take is: You cannot ignore it at almost any level, even though you hope that someone has processed the data to the point for you, that you can just use it. At least in my experience, that’s never true. You always have to go back.

Alice: We often discuss metabolomics in the context of the microbiome as the combination of the host and the microbiome pool of metabolites. And as I was preparing for this discussion with you, I was wondering since we’re talking about multi.omics: Do you know of any tools that combine several omics considering there might be more than one species at work? I’ve never seen this.

Jan: It’s a very active field of research just from the two omics side (the metabolome and the microbiome). I think that is a whole new set of methods that will be required. And maybe a good example for what we talked about earlier that there’s not always one size fits all approaches. For the microbiome with the metabolome, that’s one of the major topics. So let’s say stool samples and you have metabolites in microbes. Anyone who’s ever worked with that. And even if you don’t know much about and just picture it, it’s extremely difficult to really understand what’s going on. You have a biomass that processes metabolites interacts with the bloodstream for the gut. Somehow, some product comes out and we’re trying to interpret how it got there and which organism made it is extremely complicated and a good example for a different set of methods needed. If you want to do pathway analysis and microbes, there is an entire field, just working on the question: “Can I find the pathways that are active in certain microbes that in combination with other microbes then lead to the production of a certain metabolite in the gut that might or might not be beneficial for the human. And again, those methods are completely different than when you look at the metabolome or the proteome. It’s just something else. It’s a different biologic hypothesis. And this is why I think the answer to your question, that the method that integrates a lot of these, I don’t know if that exists or can logically exists at this point in time. I don’t even know how anything plays together here.

Alice: That’s what I expected, but I was curious to see if maybe you knew something that I hadn’t heard about.

Jan: Call me if you know anything… [laughing]

Alice: Yeah, stay in touch [laughing]. About multi omic analysis, were there other points that you wanted to bring up?

Jan: Yes. I think the manual integration of data that say into a figure or into a chart is sometimes undervalued – those pathway methods do not give you the final answer; they do not write the paper for you!

Alice: I am happy to hear you say this. I believe that but it’s really nice to hear it from different sources, too.

Jan: Absolutely. An example was the latest Alzheimer’s paper we wrote: Yes, we are going through the standard process of clean statistical analysis at large scale; then the pathway integration methods that we talked about, all of that needs to be done as a first view off what’s going on. – But our final pathway about near transmitters that we were personally interested in is a figure that my post-doc Richard worked on and no computer method could make that figure. We picked it. And, of course, you have to be careful not to make it biased and so on, but I think we shouldn’t sell what we can do manually under value. You still need the compensation part, but the final interpretation is not going to come out of an enrichment algorithm.

Alice: Do you think we´ll ever get there?

Jan: I think it will always be the case that better and better computational tools will give you more interesting views of the data, things you hadn’t expected, but it’s always more like the first page of Google for you. You still got to pick the right one and go through it and interpret it and figure it out. Only you know the field, the computer doesn’t know the context of the field and other studies in what your question is. So I think the methods will get better in condensing those complex data sets into candidates of results.

Alice: I fully agree. So then we get to the more generic questions about metabolomics and about the interpretation projects. First, do you see particular pitfalls, especially for beginners with metabolomics or people who are diving into their first interpretation projects, which they can avoid in the preparation of the data or in the actual running of the analysis that you know.

Jan: I think there’s, there’s a couple of aspects to this, first of all, and it sounds like a dry standard statement: The pre-processing really matters. I would also encourage people, students also new researchers in the field to think about pre-processing not just as this chore that you have to do. This must do step that you just want to get past as soon as possible, but it’s actually really interesting. There’s a lot of statistics; a lot of computational methods involved on the side of pre-processing the data. I know that’s not what we want to do. We want our results for our question and not work with massaging the data, but it’s absolutely necessary – You can’t skip it.

Alice: Do you also see it as a way of getting to know. This was one of the advantages of doing this manual work myself in the past is that even though I might have only shown like 5% of the data at the end of the paper, you get a knowledge of the data that is really deep. And that also points you in the right direction.

Jan: Very good point: For example, you might exclude an outlier, but if you’ve really spend some time, you know, that a sample or the particular mouse, or maybe what’s going on there; what happened. You really know it at that time and that time needs to be spent. I only know two ways: Either we spent the time and then go forward or we go forward too fast and then go back and spent the time
The challenges and pitfalls among the statistical analysis, I think what is tough in the field. And that is not even metabolomics specific. You have to be a somewhat trained statistician to run the data with all the problems that come. Can I use a method that assumes normality on normal data? Is that just some kind of mass statement that it doesn’t matter in my actual application case? Or do I really need to take care of this?

Alice: Hm. Alternatively, you can collaborate with statisticians.

Jan: Yes. And I think you must. Whenever you are running anything, even if you’re running it on the online toolboxes, unless they really take care of everything, you must have someone on board who has the understanding the statistics or is willing to acquire it, go through the tutorials, go through all these things that we might not have wonderful plots of the distribution of the data.
Again, what I really want is to go forward with my project. Right. But you have to. It can be something very small which makes a large differences in the outcome such as did I take logarithm my data. Did I scale my data before doing a PCA.

Alice: A beautiful example, do I log everything or not? Then you do your whole analysis, your whole interpretation, and then towards the end you learn that, oh, you should have done this – and you have to do everything new. This it’s quite dramatic.

Jan: Exactly. So you asked for pitfalls to avoid, so I think it’s not even possible to you name them one by one, because unfortunately you have to know how to statistically analyze quantitative data, which could also be weather data or stock data. There’s going to be similar questions, but you have to have maybe not a degree in it, but you’d have to have someone on board that has a understanding of those concepts, or you must train someone to get to that point using online resources and so on.

Alice: When it comes to the interpretation of the data, what would you say is the most time consuming step.

Jan: I prepared for this when I read your questions and that one I’ve thought about for a while. It is a good question. I came up with an answer that I think is very practical. What happens to a lot of people is that what we do a lot of times and data analysis in my opinion, is we use the tools we have; try them first and ask questions later – And then later we realized it doesn’t actually really fit our biological question. So we did a correlation analysis between metabolites and transcripts in our new multi-omics data set, because that seems logical, that’s what we do. And then when we look at the results we realized: “Was that even the question that we have asked?” Then it becomes an iterative process. We go back into our methods. So, while the execution of the scientific project should, of course, be a reiterative process. We make a hypothesis, we pick the statistical method to answer it, and then we answer it and then we write a paper. In reality, and everyone knows that, it’s a very interactive process and that is driven by interpretation. Right. It’s not just that you don’t get the results you want, but you don’t even know what to do with the results of the statistical method that you actually got. So you go back and adapt it and iterate go on. And every time a postdoc or a PhD student has to write 200 lines of code and debug it to get that new result done. – That I think is the most time-consuming step in all of it combined with the more fun part that we discussed earlier: The manual interpretation as we set the path we met that doesn’t get you to the last figure in the paper. So you must put in your own interpretation. That is time-consuming, but I think very fruitful. The iterative part where we switched between methods and hypotheses constantly, I think, is not very productive and needs been taken care of.

Alice: Do you still strive to one day, begin with your question, run in a straight line and get to the end or do you know, that this will never happen?

Jan: I have to acknowledge that while that’s what we want and it’s always easy to think about it in hindsight, in reality, that never occurs, but also in our lab and in our discussions, we’re trying to move more toward asking a question for which we picked the method instead of using the method that we know so well, use it first, and then ask the question.

Alice: This really makes sense because sometimes also you have the tools that you have, and they might be very good to answer certain questions. So why not just go for that?

Jan: We are changing our wastes a little bit. I always ask in my lab: “Who cares?” Not in an offensive way, but in an who actually cares. Right. So who’s really interested in the result that you produce. If you don’t have an answer to that and you should immediately drop everything and rethink first.

Alice: That’s a good way to go. Yes; that’s true. So your work in developing tools that requires a lot of creativity, do you also see a need for creativity in the interpretation parts?

Jan: I think yes and no. Interpreting the question also in terms of statistical and computational tools that aid in interpretation. And I think all of us who work in computational research can confirm that the most interesting new methods, the new toolbox that you publish later, all originates in some questions and some creativity. I don’t know what to do with this data. Maybe we can try this – And suddenly, for example, we had this on the networks with the MoDentify toolbox. It started as a “couldn’t you try this thing” for the PhD student and she thought about it and came up with a creative idea of how to make these modules in networks, and then it ended up being a toolbox. It is a new data interpretation toolbox that came from a creative brainstorming of how to go forward. I think therefore the answer your question, is there room for creativity in data interpretation? must be yes, because that is literally what we do.

Alice: Otherwise we end up with the old tools, because then you don’t create anything that gives you, as you said before, the beginning of the answer, but you still have to make the work of connecting it together. So we need creativity for the tools, how to apply them and then how to understand what they’re saying to us and why that makes sense.

Jan. Exactly. Especially in exploratory studies. Yeah. Where you don’t have an outcome that is clear if you know what the outcome is, there’s no creativity. You just have to pick the right statistical tests like in the cancer study. Not that they’re easy to do – but the outcome doesn’t need any creativity. For a new omics data set and you give it to a PhD student and say here, figure it out. – You definitely need creativity.

Alice: To finish, I have two straightforward questions. Which one is your favorite metabolite? And why?

Jan: Yeah, I had to laugh when I saw that question. -I don’t want to pick favorites.

Alice [laughing]: I want to love all my metabolites equally.
Jan: A colleague many years ago also said after working with metabolites (and at the time we only had a couple of hundreds) that you know, each and every one of them personally. I think for me, one of the most fascinating ones is 2-Hydroxyglutarate. It’s a metabolite that is relevant in cancer.

And why I find it fascinating is that it’s a naturally occurring compound, but with a gain of function mutation that happens to the TCA cycle in many different cancers, It suddenly makes the TCA cycle sort of spin out of control and produce masses of this metabolite is really interesting that you could change the enzyme to produce something very similar, but a little different.
And the 2-HG is being debated as one of the first onco-metabolites. So the ones where it is not just a side product off the pathogenesis of cancer that it occurs, but it contributes to it. The presence of it has epigenetic effects and contributes to cancer and the pathway spirals out of control so badly that you can actually see the metabolite even in the bloodstream elevated a hundred fold. I think that story is fascinating and I worked on it.

Alice: Thank you. This concludes our conversation about metabolomics. Thank you, Jan. And I look forward to all the wonderful new tools that you will develop for us. Thank you.

Jan: Thank you very much. And I’m looking forward to your exciting podcast.

Group leader Metabolomics at the Institute of Bioinformatics and Systems Biology (IBIS), Helmholtz Zentrum München, Germany (since 2011)

Kastenmueller group @Helmholtz

Favorite metabolite
b-citryl-L-glutamate

Discussed paper by Arnold et. al.
Sex and APOE ε4 genotype modify the Alzheimer’s disease serum metabolome

Database links discussed
AD Atlas
Omicscience
Snipa
GWAS Server

Other resources
Talk about the Omicscience platform by Prof. Claudia Langenberg
(given at the biocrates Pan-Cohort Metabolomics event 2021)

More about The Metabolomist podcast

Episode Transcript

Alice: Gabi Kastenmüller is a group leader in metabolomics at the Helmholtz Center in Munich. You’ve been working for a while on metabolomics. Would you like to add something.

Gabi: I would say maybe even a bit more than a decade in metabolomics now, and this was for me a perfect synthesis with a background in chemistry and computer science and data, but it also has to do with chemical molecules.

Alice: It converges on metabolomics quite well. From what I understand, your research focuses a lot on metabolomics, a bit on GWAS as well, mostly creating tools then to better analyze and interpret this type of data. Is it mostly bioinformatics or do you do also different things?

Gabi: I think our general driver for our research is that we want to understand what influences a human metabolome. Coming from genetics as an inborn factor of metabolism, but also influences by the microbiome, by nutrition, by exercise. We want to understand these influences on the human metabolome and the changes of the metabolome over time by using metabolomics.

Alice: And the metabolome is specifically interesting type of molecules to study in this context? Isn’t it too, because a lot of people would study proteomics or maybe transcriptomics, but you made a choice at some point to focus on metabolomics. Is it because it’s so sensitive to these different influences?

Gabi: Exactly. It’s really in the intersection of all these influences and all these influences that we know are important for developing diseases and also for treating diseases. So it’s molecular layer, very close to the phenotypes or to the symptoms; very close to what you can do against it in an easy way, like a diet and exercise, but also when you have drugs and medication, they often act on the layer of metabolism. So it’s the molecular layer where it all comes together: The genetic imprint and what we do to our bodies with our lifestyle.

Alice: In part of your work, you develop by bioinformatic tools and databases, some of which are available online. Could you tell a bit about this free access to data because people see sharing data, whether it’s metabolomics or other type of omics data as a kind of a chore, sometimes they’ll have a feeling it’s a different dynamic on your side. I have a feeling that sharing the data is part of the processes for you.

Gabi: We share the opinion that it’s an important process of work. So, as I said, the driver of our work is to understand the metabolome systemically and you won’t get to that understanding without collecting the results from very different angles of research, very different projects, where you look into diseases.

Compared to where you look into nutrition and exercise and so on. And what we try with our tools is bringing these results together. It’s not only about sharing the actual data. It’s in our opinion, also important to share this wealth of results that you get from that data. It’s much easier to share results compared to data because you don’t have all these restrictions coming from data protection.

You have to make sure that you don’t give out private sensitive data. This problem you don’t have when you are sharing results, but the essence of what you do still has to come together. Sharing the results is as important as sharing the data.

Alice: In terms of results, do you mean, for example, the associations that you find when you combine metabolomics with G was, this is the kind of results that you share, right?

Gabi: Yes. We make these online tools that you already mentioned to bring all the results, not only highlights, which you focus on in a publication, but we want to bring all the results to the people interested in these results. So if someone is interested in a single very particular gene, because he or she is working on that gene for decades and is very interested in the function there, then this person should be able to see and search easily for associations that we, for example, see with metabolites for this gene. And this can suggest new experiments for someone who is really interested in this very specific functional aspects. Usually, these type of results are somewhere hidden in supplements and the people interested in those results won’t even find them.

Alice: Yes, absolutely. And it is really useful. I’ve worked with a lot of different omics, not just metabolomics, but still I have no chance working with Gus. I’m not a bioinformatician. If the data was freely available on a server somewhere, but I still had to mine it myself. I would have no chance. The fact that you provide the associations and it’s really simple to either start from the metabolites or from the genes and to find just what came out of the analysis. And that can be an inspiration sometimes when you’re stuck, even just to try and see if there’s a direction you hadn’t thought about and if there’s something in the literature that you wouldn’t have found by looking at the general topic is it’s maybe not a mainstream theory yet. And you can find new interesting things in that way.

Gabi: That’s the goal of these tools that we try to make our association results as accessible as possible.

Alice: Do you want to name a few of the tools? We will put a list of links so people can find it, but maybe she wants to name a couple of such tools.

Gabi: In collaboration with the group from Cambridge now in Berlin, the group of Claudia Langenberg [A talk by Prof. Langenberg on omicscience can be found here], we set up a whole set of such online supplements for GWAS, which is called omicscience.org.

It’s not only about metabolites, it’s also proteins in part, but we will add further association results from big cohorts there. And there you can find also an association results from metabolites, with diseases and disease risk factors from a very large population of 11,000 participants.

Alice: These tools are still growing so that it’s not just a static result of a project or a paper, but every new study that can be added to it.

Gabi: We try to sustain these things as good as possible. It is always difficult because of funding. These kind of things are not really funded well, but we try to keep those servers up. Not only for one or two years after the publication, as a growing resource.

The other resource, which is already quite old, but still we are updating and developing this further is the snipper.org where we collect information on each and every genetic variant in the human genome. So the type of associations we link in there are results from different GWASes that are also available on other places, but also metabolite associations and associations with proteins.

Maybe a third addition: Our newest one, the AD Atlas. We really try to not only make single associations accessible one by one (if people are interested coming from a different metabolite or gene) and just give a list. We focus in the AD Atlas on bringing these lists and these associations together into one network to get a more systemic view on the interactions between the different molecular layers between the different disease phenotypes. And in this case, we made that specifically focusing on Alzheimer’s disease related phenotypes. In general, this is not restricted to the disease when it comes to the molecular backbone. We use really broad data from healthy big cohorts – A lot of the gene-metabolite association analysis or gene-protein association analysis.

Alice: So you combine GWAS – metabolomics – proteomics. Other things as well?

Gabi: As it’s specifically focusing on AD we are also including metabolite-disease phenotype associations, and also tissue specific expression profiles, for example, that are particularly interesting for AD. So we have different brain tissue and gene expression information from a big US consortium working on AD and that’s all brought together in this database.

Alice: Did you find new ideas about the pathology of AD from this association that you make now in these networks?

Gabi: So far, we mainly looked into it in a more explorative way of starting from specific group of metabolites that we saw associated, or a set of genes that is interesting from a particular perspective, like targets of a drug such as statins. And then, different statins have different targets beside the main target.

It is starting from questions like these and then we create and analyze sub-networks in terms of enrichment of genes we have in the bigger network. From these enrichments, we got to suggest a drug repurposing by combining it with different databases where you have this information on drug screens.

That’s what we already tried to show that this network is of value. In the next step we want to really mine the complete networks. That’s what you were saying. We just want the data to speak to us. So this will be the next step that we provide the possibility to analyze this huge network of millions of data points.

With graph based methods.

Alice: So at the moment, it’s not yet available to the public to use, but this is the plan.

Gabi: What people can do with the Atlas already now is to explore starting from a set of genes set of phenotypes, et cetera.

Alice: So now you can already filter and play around and see what comes up – it’s already available.

Gabi: Yes. It’s already available and you can filter, you can search, but the step that is missing and that we want to go now is to take the full network that we created and mine it in a more holistic way.

Alice: Are you using Alzheimer’s now as a kind of training disease and then planning to maybe use the same strategy that you hold now with this disease to other interesting diseases, is that the plan?

Gabi: The backbone of molecular interactions will be the same anyway, because that’s coming from the very big, more or less healthy cohorts, right? And we plan to couple this with different disease specific databases, information, or consortia data in the future to make different atlases for different diseases, but also to use that for a better understanding of co-morbidity – That is also one of the interests of my group. We want to understand on the metabolic level why obesity and the linked metabolic pathways, a risk factor to basically all age-related diseases. Why is it going to type two diabetes for some individuals and to Alzheimer’s for other individuals. These pathways are all connected and this is known, and we want to understand how.

Alice: That’s a lot of fascinating work ahead. I think.

To prepare this interview, I put your name in PubMed to see what comes up. And what was interesting with you is that if looking at the first two pages, you have a multitude of very different papers that each one seems fascinating, but there’s one paper that I find particularly interesting because it talks about a topic I’m very interested in is the sex differences and metabolomics. In 2020, there was a paper where Matthias Arnold was first author and you were last author that talks about sex differences. The paper is called – Sex and APOE ε4 genotype modify the Alzheimer’s disease serum metabolome

Firstly, it clearly puts forward the idea that sex differences are a topic in metabolomics and that we shouldn’t avoid it. I also like the paper because it had two main figures which are radically different. So the first figure is a highly complex visualization of the data, really a lot of information. I don’t know if it’s a classical way of showing the data, but for me, it seemed to be very creative to really show the pulling of the, of the male and female differences and metabolome together with other factors and then managing to extract the few metabolites that seem to be really strongly influenced by it by sex.

And so a lot of information condensed in the picture. And then the second figure, the complete opposite, very simple box plots that has the most beautiful message I’ve ever seen. I don’t know if you remember that figure.

And I found the contrast between those two figures. Really interesting. So just to describe it quickly to the people who are listening. The second figure is a box plots of the levels of proline in the patients in blood. Without doing any special stratification of the data, when you look at the proline levels compared in two groups, one that would be diagnosed with Alzheimer’s and one that wouldn’t there’s no, there’s no difference visible.

And then you start stratifying based on sex or ApoE status. And again – not much to see: You see differences between the two sexes, but you don’t see that there’s a difference in Alzheimer’s. Then the results of the twofold stratification and suddenly it happens. For males – not the most relevant information, but for females, you clearly see that there’s a difference between Alzheimer’s and control and that organizing the data in this way allows to just reveal this important difference.

Would you like to comment on what this figure means and what the, the overall topic of sex differences in metabolomics is about?

Gabi: We all know that there is a huge sex difference between metabolite levels, but usually people think that these differences are a different layer. In case of disease or treatment, people then think that on these different layers the effect [of the disease] in the different sexes is basically the same. That might be true for most of the cases. We are all functioning in a similar way, biochemically from the, from the type of reaction, right. Even though the levels can be very different, but this example shows that some of the effects in Alzheimer’s disease that we see metabolically is only relevant for a small subgroup. So you have to have the genetic risk and be female that you see something in proline. That’s a very simple message. And that was what we wanted to convey with this figure, because it’s otherwise from the analysis, it’s sometimes a bit hard to explain why are we looking into these subgroups at all?

Because – on average -, we haven’t seen any difference in proline. So why were we looking into that more deeply – that was what we wanted to show.

Alice: It is really shown nicely in that figure because the figure is so simple. Sometimes you need the complex visualization, like in the first figure, to show as much as possible of the data and sometimes it is really important, like in the second figure the specific message is the simplicity.

Gabi: The first to figure out the outcome of the whole analysis, right? So this does not only happen for proline. You have different effects also for other subgroups and other combinations. And you also have these homogeneous effects that are really the same for everybody. Apparently.

Alice: And this is important than when you’re interested in future treatments and which targets to choose.

Gabi: Yeah. And it’s also important to know about the heterogeneity of effects because those associations can also come up in unstratified analysis if you happen that your cohort is a more biased towards a specific genotype or any other reason for this kind of bias.

And then this association can come up overall, but it’s indeed it’s only relevant for a subgroup. And maybe, if you target the other groups can; it can be harmful to target this. Right. So we don’t know. And that’s why we wanted to show that we have to analyze the heterogeneity and not only say, oh, we see a particular global effect or p-value.

Right. And we can replicate it. Maybe that’s it. It’s not the end of the story. We have to look into subgroups more carefully, especially in diseases where we already know that we have phenotypic heterogeneity even, right?

Alice: Yes. And this brings also the question of how to find new markers or new targets because here in this example, the way you stratify is based on known risk factors. So you know what to look for and you know how to stratify. Sometimes there might be a differential effect between different subgroups that we don’t know about because we simply don’t know that the subgroups exist. Like we don’t know that something is a risk factor because it hasn’t been discovered yet.

Gabi: So that was at the end of working on this paper exactly something that we cut out for the new analysis that we go away from known risk factors to a stratification using the metabolomics data. So I’m approaching things from the other end questioning whether we can use part of metabolomic status, specific metabolites, subgroup, individuals, and then in a second step, check whether we see associations with AD phenotypes and that’s something we are currently putting together in another publication where we use this approach of sub-grouping using the metabolomic state.

Alice: And this is one of the powers of metabolomics! Would you say that there is a similar kind of hope now with metabolomics there was with genomics because 20 years ago, genomics was going to solve biology, kind of, and there are lots of things that were found out and there are lots of things that are still being done with genomics, but there are lots of things that we now know don’t just rely on our genetic code or even how it’s regulated at the epigenomic level that there are things that happen directly at the metabolite level or that are visible later at the metabolite level. Do you see a new kind of hope coming from metabolomics?

Gabi: Yes. I see the hope that we can get information on different parts of lifestyle and genetic influences that come together in an individual directly with one measurement that’s I think the promise and the hope of metabolomics. I think we are getting close to what makes it always very complicated is that we have this fluctuation of metabolism all the time. So anything will influence the metabolome. Be it what you ate two days ago, or whether you did a lot of exercise the week before it all will influence your current metabolome. I think there are two solutions. One is to have a good knowledge of this fluctuation. By looking into time resolved data. That’s another branch of the group where we are very interested in understanding when do we see which metabolites are affected by these things, which ones are not. To get a really good knowledge about this [time resolution] is one solution, in my opinion, the other good news is that even if we use metabolomics data from all these different points during the day and all these different challenges, like exercise, different types of nutrition, if you take it all together, we have seen that the metabolomes of an individual is a very stable thing. Even over years, we have shown that.

Not in each and every metabolite, but altogether people stay closer to themselves than to other people closer to yourself no matter what you do to your metabolism. The other thing that we saw in the analysis of bigger cohorts over years (and the changes over years) is when the whole metabolism changes dramatically in an individual: That’s a bad sign. That’s an alarm sign. Yes. And I think that is an argument for using metabolomics as a monitoring tool to see when things are disbalanced in a way that the individual body cannot cope with it anymore.

Alice: You could imagine something like getting the metabolite panel checks every year, just so you can catch things early; and then after 10 years you have 10 pictures of your metabolome and you might see an outlier when you start developing relatively early on a disease, that’s might get caught earlier that way.

Gabi: That’s a helpful, when you look at the complete metabolome as such as a combination of metabolites, it gives you an idea, but also you have the different metabolites where you might be in the normal range still when you compare it to a normal range (as it’s done with the classical clinical chemistry). When you have your [own] monitored levels, you can see a trend already much earlier. Maybe you have always been very low in that level and high levels are bad, but you have a clear trend rising. You have a much more relevant information. And that’s where I would hope that we get metabolomics into the clinic as a supporting tool.

Alice: There’s much that can be seen through it. I mean, it’s used in certain applications in the clinics especially with newborn screening, but there’s really a lot more that could be done if metabolomics enter the clinics.

Gabi: The beauty of metabolomics I would see that we can also label specific metabolic classes or specific metabolites, a combination of them to modifiable risks that can be related to your lifestyle or to the microbiome. We have the information, for example, maybe this comes from a disbalance in your microbiome.

Maybe you had antibiotic treatment, too many of those, or something similar. So we can also see all the metabolites that show up when you look into liver problems. You see those combined with those indicating microbiome dysbiosis and act on this. I think their metabolomics has a lot to give.

Alice: Yeah, I think so too. I agree. I like the idea of lipids being more than just a membrane components, but being active members of the metabolome and of our biology. What’s your view on this?

Gabi: I’m not a particular expert on Lipids, but what I noticed over the years is that lipids come up as associated with diseases almost everywhere. Any disease to look at a lipid is involved as well. I’m pretty sure that lipids are much more than just things that you need for the membranes – it must be because they are so diverse. Why should nature make all these different things? What makes it very difficult in my opinion, when doing these lipid screens, is that the classic biochemical functional information on each and every particular liquid that can be measured nowadays is missing. So no one has done the hard work of finding out what this particular lengths, there and there the double bonds, all these [chemical functional] things that we can imagine now is really doing and in comparison to the next length and the next double bonds.

All this biochemical hard work is missing in the literature, and that makes it really difficult to do interpretation. And the other thought I always have, when I see lipids in many cases: I have seen really bad things happening in terms of interpretation. When people don’t know what the labeling of this measurement means.

Alice: What do you mean the labeling? Like, they’re not sure of which lipid they’re looking at?

Gabi: In lipidomics we have these different layers of precision. The lipid structure, but you don’t know exactly what the fatty acid chains are behind the measurement is. For example the label PC [Phosphatidylcholine] something it is hard to really capture what’s behind that label and find the interpretation and literature on that, because literature is more there for very specific ones, very specific lipids, very detailed things but what you measure is much more of a bag of things.

Alice: And in a nomenclature that is used in the publication might be different from the one you’re using as well, which makes it difficult to find information.

Gabi: Exactly. I think this complexity of lipidomic measurements is not appreciated.

Alice: I think that’s also why when we see a beautiful lipidomic study with a beautiful story of what the lipids are doing, why it’s so fascinating and exciting because some of the most interesting metabolomics or lipidomics papers that are read sometimes are really about lipids because you see it’s a whole new world, a new dimension that opens every time there is a cool story that comes out about it. And also because there’s this scarcity of information, both on the function and the structure of what is measured, that makes it difficult for the interpretation. Speaking of interpretation, this is the last topic I want to address more specifically, though this whole podcast is about interpretation, of course. I explained to you a bit when we spoke before the picture I have about interpretation project, where you start planning your project and then you have execution steps where you prepare your data and operate the tools that you want to use, and then check that everything is fine and finally, you can do the interpretation work so that you found in your case, for example, you found the associations that are significant between metabolites and genes or this kind of things, your confidence, you did a good work, and now you can do the interpretation and try to figure out the story behind let’s say Alzheimer’s or whatever the topic of the paper is.

In your experience, what would you say is the most time-consuming step of the process? What’s your feeling about it?

Gabi: There are two steps that at least in the projects we were involved in always took most of the time. The first step in getting to know the data together with the experiment. That´s a very time consuming step, because no matter whether you come from the experiments, so you are the one who does the experiment and have the idea about the research question or whether you come more from the metabolomics and you know, the metabolites; it’s always hard to really capture the rest. So for the experimentalists, who has a clear idea of what the research question is often has a very hard time to capture what’s in this metabolomics data. How much can I rely on this one? What is that metabolite?

What does it tell me? And from the other end, when you’re more the metabolomics expert, first have to understand what the research question is in a very detailed way to really help the people to make most out of the metabolomics data. And no matter on which side you are; all the studies we did were highly collaborative. You always have different parties involved.

Alice: Yes. In fairness with metabolomics, I’ve spoken with some people who do the planning, the measurements, the analysis [themselves] but I think these people are quite rare and what more often than not it’s a group effort.

Gabi: It’s possible but everybody has to make this common crown somehow.

Alice: And you have to find a language you both understand, which is sometimes difficult. There are even words that are used by biologists and informaticians or chemists they’re the same words used for different things like “feature”, for example. I always find it complicated. Even “interpretation”, some people think of interpretation something like interpreting the signals of my measurements to know what the metabolite is when other people like me from a biological point of view, say, I’m going to interpret now what’s going on the biology level. You can have conversations with people for a while until you realize you’re not talking about the same thing.

Gabi: I have this actually with a chemist and a computer scientist, they were discussing upon putting together a manuscript. I would say for 20 minutes, they were discussing a graph and after 20 minutes, it turned out that the whole time the computer scientist was thinking of the graph in a very mathematically well-defined way of a network that wanted to set up for this publication and the chemist was just talking about a conceptual figure for the paper – that things happen, at least in studies we were involved in. It takes time and I think we should give ourselves this time to really understand what we are doing together.

Alice: Yes. Because the risk, if you don’t do this, is that you go further down the steps and then, when you realize [you did not have a common understanding] you have to go back to the beginning. This is in my sense the worst, because it’s absolutely normal and you should take the time to prepare and plan and make sure that everyone understands what they have to do for any project.

Gabi: Especially in big projects, even in the funding schemes, they always expect the best experts in each and every discipline. So the best statistician doing new algorithms or coming up with new ideas to analyze and the best analytical chemist, but they are not bringing them together. They sometimes expect that one part is done only by the analytical chemist. [The other part] is done only by a statistician and that’s not necessarily the best outcome for the research question. You clearly need these experts, right? But you also need those that bring things together and that’s sometimes a bit forgotten.

This is important. So that’s one part of the interpretation where I see it’s time-consuming. And the other part is bringing together current knowledge in literature from other current screens with what my results are. That’s also a very difficult time consuming thing because you have, sometimes you have to read deep into the publications.

Alice: And this is more often the one person job, isn’t it. Or do you manage to do this in the group? Because this is much more difficult to organize in the group.

Gabi: That’s true. At least you need one person who wants to dig into it deeply. Without that person it usually is very difficult because it stays very shallow and together in one head first, then of course the discussion is needed with all the experts. Sometimes you find something and you think it’s important because you are not so familiar with that particular field. So you need that one person who is willing to connect all aspects in her head first.

Alice: Then once you’ve made your own connection of the dots, you need to go out and also expose your theory to the world as you can’t be an expert in everything or an expert in every biological field. So you need to have then inputs to see if that makes sense or if maybe you need to rework parts of the story.

Speaking of this interpretation part from the biological point of view, where do you see a place for creativity? Is it there? Is there also creativity involved in the previous steps for you or everywhere along the way, from an original scientific question to how to do the work to also how to form the team? What’s the place of creativity?

Gabi: I would see it in each and every step. I think that would be the ideal case. Of course, there is a lot of need for creativity in that last step where maybe you can have more standardized processes before, but if you really want to get the most out of the story, I think creativity at each step is bringing things forward.

Alice: My last question for you is what is your favorite metabolite and why?

Gabi: Yeah, I saw this question. Funny one. It made me saying it’s changing with the studies. I only can say what my current favorite metabolite is. I would say it is beta-citryl glutamate. Why? Of course it has to do with the current project we did where we had muscle biopsies from people doing resistance exercise. That’s a collaboration with exercise biologists from Cologne and Munich. It was not the first time that I saw beta-citryl glutamate in this study, but there, we saw a change in muscle for this metabolite. And I think the metabolite is a very good example, why I liked to do metabolomics instead of very targeted things only (like only clinical chemistry). This is a metabolite known for a while, but it has been described mostly in embryos for development of the brain and later has been seen also in Spermatogenesis, but, apart from that, it hasn’t been described much. Metabolite association studies with different phenotypes, mostly in exercise – It comes up. It’s basically nowhere described as key point or a highlight because nobody knows what the thing is doing. I found it very interesting because I immediately looked up almost automatically: Do we have a genetic association with that? Sure. If we do, I could have found that in knowledge bases as well, because there is a particular enzyme dealing with.

Synthesizing this, which always tells you that it must have an important role somewhere, and nobody knows where. And I think there metabolomics can help a lot to look into areas and fields where this metabolite might be of importance. We can get to the next steps or suggest experiments to see what the role of this metabolite is.

Alice: It’s a beautiful example of both types of applications of metabolomics as well. You might find new markers of a disease state, or of exercise or excessive exercise, but you might also better understand muscle physiology by looking into a new pathway that’s that no one looked into before in that context. It´s really cool. That was a great example.

I have the same. I always change my mind just depending on what I’m looking into. Is there something else you would like to discuss?

Gabi: I talked about all my favorite parts – time resolved metabolomics, different metabolic challenges like exercise and nutrition, the stability of metabolomes over time. On the other hand, not only for the fluctuations, but also the stability and also my favorite topic of making our association results and all the information we have about metabolites accessible to others in the way that they don’t need to be bioinformaticians.

Alice: On this beautiful note, I would like to thank you for joining me on the podcast. It was lovely to discuss with you.

Gabi: Thank you very much for the discussion. That was a great pleasure.