 So congratulations everyone. You've made it to the very end of the workshop Also, what a congratulate Rashad for doing such a great job of orchestrating things and making sure we all Were online and in our chat rooms when we needed to be I think Jeff and our TAs have done a superb job of Teaching about metabolism. This is certainly the most popular part of the workshop We try and give people a fair bit of time just to sort of play around and every year there's something new And and more expansive in metabolism analyst. So I learned a lot just by sitting back and watching what Jeff and the team have done Today, I'm just gonna try and do something quick before all of you fall asleep And before you get too hungry But this is to look at about the future of metabolomics It's a little speculative And so I'm certainly inviting comments or questions But it's based on Being involved in metabolomics for a long time, but also being involved with a lot of groups around the world Doing some pretty cutting-edge work in bio in metabolomics So most of you are taking this workshop because you're interested in it and most of it because you're realizing that metabolomics is growing in popularity When I started in the field of metabolomics back in 1998 It didn't even have a name We were choosing all kinds of different things. There was only one or two papers published a year As of 2019 there were more than 6,000 papers published in the field and it's growing pretty consistently The worry is Where is metabolomics going is it kind of reaching a peak right now? And you know next year everyone is going to leave and find something else Maybe everyone jumps into the microbiome or someone finds something more in epigenetics Is it going to level off? Which is sort of the way that most things have happened in things like proteomics in some fields of genomics Whereas again that continue to to climb and how much longer will it climb in terms of popularity and interest? When you look at where it could go you have to try and identify where there's some essential bottlenecks in metabolomics Some of these bottlenecks are things that we brought up already One is the fact that metabolomics is pretty labor-intensive whether it's preparing the samples running the samples or even doing the data analysis Another point we've highlighted is that often we're happy if we just get 50 to 100 metabolites when we know that in fact there are hundreds of thousands so we have very incomplete coverage and that's been a real sore point for metabolomics We know the equipment is very large expensive We've also highlighted the challenges of lacking quantification in fact proteomics Basically hit the wall and declined quite substantially in interest because it never really got around the problem of quantification We also have a real challenge in translating metabolomics discoveries either into the clinic into the veterinary classes or approaches into industry Into environmental applications. So the translation from the bench to Real-life applications is challenging and then a lot of money Has driven You know our activities in the omics world Largely sponsored by pharmaceutical companies It was big pharma that pushed genomics. It was big pharma that pushed proteomics It was big pharma that actually started metabolomics. It's big pharma that's been pushing precision medicine. It's big pharma that pushed Big data for medicine So if you can make your field matter to big pharma, then it also helps to Advance the field So I'm going to talk about some trends And I'll also discuss some aspirational thoughts about these trends in metabolomics I'm going to talk about automation I'm going to be talking about efforts to expand metabolome coverage and be talking about Quantification and the importance of that and where people are going I'm going to take some examples of how metabolomics can move from the lab to the clinic and how metabolomics is Coming back into drug development and discovery and becoming more relevant So automated metabolomics is becoming more and more ubiquitous I'd mentioned a couple of these things like Brooker's efforts in NMR to do lipoprotein juice and wine Where it's possible process samples in about five minutes automatically I'd mentioned the software tool called Kinomics some of you guys have used it, but it has a semi automatic method Science has introduced a lipidizer, which allows you to measure Fairly automatically about 1100 lipids and biocrities has been a company in Austria That's been offering kits to do targeted metabolomics first the P 180 then the P 400 then the P 500 And there are other companies as well that are starting to offer these automatic or semi automatic approaches to doing metabolomics They sort of either open the kit or press the button or load the system and you can do metabolomics while you sleep There also a growing number of service providers that also allow you to make metabolomics pretty much automatic too So you don't have to buy a mass spec. You don't have to be Someone who's spending all their time learning the techniques, although it's good to learn them So there are core facilities. I've mentioned the metabolomics innovation center and that's where My and mark currently working that's also where Jeff cut his teeth when he was starting off as a Graduate student in a postdoc their company's like metabolite There's about a half dozen Metabolomic centers in the US. There's the National Phenome Center, which is a network spreading across in Singapore Birmingham and London There's Nightingale a Finnish company that does metabolomics. It's running through millions of samples in the UK Biobank Netherlands has a core facility. These are all places around the world that do Very extensive very high quality metabolomics Measurements and so you just have to send your samples off to them and a few weeks later the answers arrive on your doorstep Of course, it's not free, but these are basically Operated as an at-cost service to to the public We've already seen some of the tools Basil was one that you guys have already used that's essentially automatic NMR Metabolomics you've seen GC auto fit also trend towards automated GC MS metabolomics and then as the kits have evolved the automated LC MS metabolomics So the idea is if if you can get These approaches more automated whether it's NMR whether it's GC MS or targeted LC MS It will make metabolomics much easier for you guys much more accessible for your colleagues And much more useful Right now the real software challenge is trying to automate Untargeted metabolomics and you guys are seeing some efforts with that that Jeff is working towards through Metabolicalist are but it is really really challenging given the diversity of preparations samples and analytical methods But that is an aspirational goal that we would like to see in the future Expanding metabolism coverage This is the Achilles heel to metabolomics and if it's not solved soon, I think metabolomics will become largely irrelevant So in terms of untargeted metabolomics Typically only about 2% of the detected MS peaks are actually identified Most metabolomic studies report fewer than a hundred identified metabolites and yet we know there are hundreds of thousands of known metabolites Most metabolomic studies don't even reach MSI level one many failed even reached level two And in terms of coverage compared to genomics and proteomics We're at roughly one to two percent of what is possible with these other technologies Whether it's clinicians or environmental chemists or toxicologists They generally lack tool trust in the tools that don't have broad coverage and metabolomics really as yet does not have broad coverage And we talked about the the different levels of sensitivity of the tools NMR GCMS and LCMS and we know that the less sensitive methods actually tell us more about the metabolome than the more sensitive methods That's because in things like NMR GCMS. You're detecting High abundance compounds that we know a lot about whereas at the LCMS we're detecting very low abundance compounds that we know almost nothing about It's a challenge so we'd like to be able to go to more known unknowns in LCMS and as yet we haven't been able to figure out how to do that So what are these known unknowns? In many cases we have a really good idea about the compounds that are in the environment We have a good idea about the compounds in our foods A good idea about the compounds that are in our bodies But all of those things whether they're foods or pollutants or even the endogenous compounds Go through something called bio transformation or chemo transformation So these known compounds get metabolized and they're converted to mostly unknown and uncharacterized compounds And so this is called the chemical dark matter in metabolomics and we currently estimate that the number of Unknown unknowns in the universe at least the environment that we live on earth is about five million That's a huge number And you've seen this slide before as if we had to try and Synthesize all these compounds even though we don't know what they are or if we try to Plug them through artificial guts or artificial livers or artificial transformations It would cost billions and billions of dollars And that's not possible. No government's willing to to make that kind of investment certainly not now So the approach is I think that we're left with is to see if we can come up with ways of generating those structures computationally and Generating their Spectra computationally and that is something that is certainly more feasible So this is given rise to something called in silico metabolomics and this is an emerging trend that's happening around the world It's making use of computers and it's making use of what we know So we've talked about the compounds that are in drug bank or HMDB or others And those when you add them all up there about 250,000 compounds if you run them through this tool called bio transformer it does the Chemical predictions it imagines these things in the liver it imagines them in the gut And it and it does essentially chemical prediction it predicts chemical reactions And right now the estimate is that roughly each compound can produce around 20 different metabolites So these are metabolites and metabolites So that's the assumption and there's good reason to believe that that actually represents a good portion of the unknown or the dark Metabolome so now what do you do when you've generated all these structures? Well, once you've got those feasible chemically or biochemically feasible structures Then you want to try and generate their observables because in metabolomics. We actually don't see the structures We see their spectra. We see their mass spectra their NMR spectra. We see their retention times So what we want to do is be able to predict those observables the spectra ms ms NMR gcms the collisional cross-section and We have to do that for literally hundreds of thousands not millions of compounds But the result is that we now hopefully would have not only the structures predicted or known as well as the predicted Observables or known and from there we can use the the rules and methods that we talked about before about how you do database matching To figure out what your unknowns are So it's a it's a technique in silicon metabolomics it uses prediction of biologically feasible metabolites It uses prediction of observables from retention indices retention times NMR collisional cross-section Molecular weights and ms spectra. The tool I mentioned is called biotransformer. It was developed by a student of mine Yannick Who's now working in the US? with And It's outlined in terms of what it does that uses a range of techniques or tools To take a structure run it through a bunch of reasoning engines and To determine what would happen and if it was in Phase one metabolism that's cytotrome p450 space to metabolism, which is what happens also in the liver Glucuronidation What happens when compounds are in the gut what happens when compounds are out in the environment? Or what happens when compounds are just floating around and exposed to a variety of enzymes, which are not always perfect in their specificity So It's this database which uses known handmade rules as well as machine learned rules has been created It's been tested that seems to reform quite a bit better than commercial tools. It has very good precision and recall So it came out last year and it's being used by quite a few people around the world There's a website It's not as good as we'd like it to be but it's pretty quick It allows you to take a couple of structures at a time and to predict the products and reaction mechanisms And it's now being run to generate millions of compounds and we hope to have it finished in a couple months We've been testing it as well as someone was taking Green tea metabolites giving them to rats collecting their urine and when you do that you often see many many unknowns and so using Biotransformer they took these green tea polyphenols and had biotransformer predict About 30 or 40 different possible structures and they were able to identify 22 of them And they were able to suggest at least 12 novel structures that they're trying to confirm right now So it allowed them to identify many more compounds and they had even hoped to see And it in top of that it was a little able to essentially generate hypotheses about a lot of the unknowns or unknown unknowns That they couldn't figure out so The utility of in silica metabolomics depends on having not only biologically transformed or biologically feasible structures You also have to be able to accurately predict those properties mass spectral properties and Other chemical properties so we know that if you've got a structure you can calculate both molecular weight molecular formula We can calculate approximately retention time We can calculate GC retention indices quite accurately Axia aphia has come up with a very powerful method for doing that As we've learned about CFM ID we can calculate MS spectra fairly accurately and a Mars spectra are also quite accurately calculated infrared and Collisional cross sections can also be calculated very accurately So we talked about CFM ID and how it's steadily improving in terms of its performance Both using machine learning as well as hand-made rules, but there's also great interest in measuring Ion mobility spectroscopy and calculating collisional cross sections And there's a number of instruments now are being sold all the major manufacturers have them Which allow you to predict or actually you can measure collisional cross section But now there are ways to predict them and what's quite surprising is how accurate the collision cross section can be predicted and how it's dependent on different Adducts and the types of addicts that are produced will lead to different collisional cross sections and using things like machine learning or a mix of quantum mechanics and machine learning The error that people are able to get is down to about 3% or less And this is in some cases sufficient to distinguish between isomers. It's sufficient to distinguish between Compounds that we normally can't distinguish And so collision cross section is a very promising route to Identify these unknown unknowns and to do sort of standards free metabolomics Because if you can predict the mass or you can produce the max best spectrum where the collision cross section it narrows things down quite a bit The other thing is that NMR is still one of the most valuable and useful tools for identifying and Characterizing novel or unknown compounds almost anyone who does natural product work or anyone who's ever done novel compound Characterization has always had to return to NMR There are a number of commercial programs that allow you to predict NMR spectra the pretty quick problem is that None of them are free and this has always been a problem in chem informatics. So We've been spending a fair bit of time trying to develop a freeware program that will allow you to accurately predict The spectra and the chemical shifts for any compound in any solvent proton carbon 1d 2d and also predict not only the chemical shifts, but all the couplings and the prokaira ones We've tested this against some of the best commercial ones and the program itself does about 20% better than the best commercial ones And it does almost as well as a very very accurate quantum mechanical ones which take hours and hours, but this one is very fast just a few seconds so there's a clear trend and I think exciting possibilities where advances in machine learning and deep learning improvements in machine precision And how we can measure things are allowing us to accurately predict a whole range of spectral parameters Which opens the door to this idea of in silico metabolomics And it's not just a pipe drink. There's a couple of papers that appeared a few years ago one by Lloyd Sumner on the right Are they predicted a whole bunch of? metabolites biologically feasible ones or Another paper published by group in Switzerland doing natural product chemistry where they used CFM ID to predict a whole range of spectra and use those to identify a large number of previously unknown compounds So it works It's worked in our hands. It's worked in other hands. So this concept of in silico or reference free metabolomics is is feasible Certainly if we can make mass spectral prediction and close the cross-section in a more spectra even a little more accurate. This would be be a huge win and Whether it's predicting the metabolites more accurately would also be a huge win and I think there's still Efforts abroad and also in our labs and other labs to prove that in silico reference from metabolomics can work consistently If it does then the limitations that have been holding us back in terms of metabolite comprehensiveness I think will be cleared somewhat quickly Quantification is something that I've been harping out for a while and it is a trend as well I know many of you are doing or working with Untargeted metabolomics and that's certainly been a theme for today But it is an issue that's been brought up over and over again. So even now 90% of papers in metabolomics are Untargeted semi quantitative less than 10% actually use absolute quantification But if that trend continues metabolomics going to be in a lot of trouble And so will many of you in terms of your career path Because in order to translate your findings into something useful you have to have Quantitation and this is the thing that killed proteomics Proteomics never got to be quantitative and as a result most proteomics labs have now merged or evolved to become metabolomics labs Because no one was able to convert proteomics applications So there are a number of not only automated but also quantitative Platforms that are available. I've mentioned the broker one. I've mentioned biocrities. I've mentioned kinomics These are commercial organizations that sell or produce or support fully quantitative metabolomics Now Historically targeted or quantitative metabolomics was was not as exciting or as impressive as untargeted So in the even just until a couple years ago people were generally happy to get 50 Top to maybe 150 metabolites identified through targeted methods Whereas with untargeted methods with a fair bit of effort people could get up to like 300 or 400 But over the last two years Companies like libraries have released other kits. So they now have the P 500 kit which measures up to 630 metabolites That is consistently more than what most people can measure using untargeted metabolomics Metabolon has also evolved its platform and they can routinely measure more than 600 metabolites identify them they can at least Annotate not positively identify about 1100 metabolites and their quantitation is pretty good So Quantifiable methods with metabolon commercial methods like biocrities Those are actually in most cases beating the best that untargeted metabolomics groups can do Likewise, if you go to lipid approaches sci-ex lipidizer, and then there's other Hand-made approaches developed in Australia where people are routinely identifying up to 2,000 lipid species and semi quantifying them quite accurately We've talked about some of the quantitative tools that are available for automatic NMR and for GCMS. These don't give you the high numbers that LCMS do, but they certainly make it fast and quite reliable So in terms of quantitative targeted metabolomics on average, it's 10 to 100 times faster than untargeted metabolomics So you guys should consider that when you're thinking about your experimental design Typically now with the latest techniques coverage and quantitative metabolomics is either reaching or exceeding Routinely what's possible with untargeted metabolomics? So that's an important transition That I think you need to consider There are also some new methods that are being developed that allow you to do quantitation without isotopic standards Looking at ionization efficiency using machine learning techniques And these are pioneered by a group in Estonia, and they've been shown to work in a wide range of applications The other thing to remember is that if you're trying to merge metabolomics data from different platforms say NMR, GCMS IR, UV Or if you're trying to get it from different labs, the only way you can merge your data is if you have quantitative data It will never happen if you use untargeted metabolomics data. It's just impossible So these are things to remember when you're thinking about What sort of experiments you should be performing? And this is being realized by many more senior labs and core labs around the world that are doing metabolomics So right now routine quantification by targeted metabolomics of a hundred hundred fifty metabolites is is is there Up to five and six hundred through commercial labs or through work that many labs are working on internally is now possible through Commercial groups like metabolon. It's possible to get up to a thousand for specialized lipidomics. It's well very easy to get over a thousand metabolites If we were to use ionization efficiency and machine learning and other predictive techniques that we've talked about Could we get up to a thousand five thousand metabolites? I think we can And and this as I say would largely be a form of targeted quantitative metabolomics So those are some key bottlenecks Making metabolomics cheaper faster better more comprehensive But how do you make metabolomics relevant? And I think there's a couple areas where metabolomics needs to go and if you can make it More quantifiable if you can make it more automatic if you can make it cheaper and better Then it can move from the lab into the clinic You guys learned a fair bit today about biomarkers and rock cures and the statistics and methods that Metabolanalyst uses Over the last 50 years. There've been more than 750,000 biomarker papers published in PubMed That's a pretty substantial portion of all of the publications But of those nearly a million biomarker papers less than 250 have actually been approved for clinical use So that's a terrible batting average Right now even though proteomics is celebrating its 35th anniversary There isn't a single biomarker for proteomics that's been approved And when you think about gene chips and transcriptomics, there's only been five biomarker tests that have been approved And the reason why Transcriptomics and proteomics are doing so bad is because they don't do quantization The other thing that many of you probably don't know is that almost all of you have had a metabolomics test Um, probably only a few of you have had a genetics test But all of you, uh, if you're under the age of 25 or 30 have had a newborn screening test When you were born within the first few hours, uh, they took a blood spot They did a heel prick and they collected blood and they sent it into a mass spectrometer And they looked to see if your blood had any unusual metabolites So there are now literally millions of people around the world who have had metabolomics tests It's just you don't remember it because you were a baby Your parents don't know it because it's done without their consent Um, and no one calls a newborn screening metabolomics But it is So if you think about what the status is in terms of metabolite or chemical testing, um, it's pretty widespread the number of approved clinical chemistry tests or metabolomics tests Uh in in this case western canada is over 300 Uh, the number of approved genetics tests, uh, in western canada is about 130 the number of approved clinical protein test is a little over 100 And then I mentioned the statistics about transcriptomics and proteomics tests So in fact metabolomics already is in the clinic and it's doing better than other omics, uh, techniques Why is it so popular and even though we don't even know that it's popular It's because uh metabolites are great at measuring phenotypes Um, these are some studies that we've been working on the last few years We've looked at how you can use metabolites to predict diseases So this is taking someone who's apparently healthy taking their blood or urine and saying well in a few months to a few years You're going to develop this disease So it turns out it's very accurate for predicting preeclampsia. This is high blood pressure that pregnant women develop So at the first trimester when they're very healthy seemingly normal if you take a blood sample from the mother You can predict which mothers are going to develop preeclampsia or not This opens the possibility of having prophylactic treatment You can also detect in pregnant mothers whether the infant or fetus has a congenital heart defect Again in the first trimester trisomy 18 or trisomy 21 again in the first trimester So not only can you use metabolomics to predict what's going to happen with the fetus in a mother You can also use it to predict what's going to happen to someone who's been newly diagnosed with cancer Now most people don't necessarily buy die of the cancer itself They die from the effects of cancer and namely that's catechia or muscle wasting And if you know of someone who's had cancer or died from cancer, this is this is the prime Killer it's it's the wasting that develops and so if you can predict which people are going to be susceptible to catechia Then you can start making interventions Particularly in nutrition and it turns out just by measuring a person's urine You can predict who is going to be susceptible to catechia or not Now predicting diseases is one thing Diagnosing diseases is another and so we've been using metabolomics looking at blood and urine to diagnose Rejection so you can instead of taking a kidney biopsy which is used with a giant needle where they stick it on your back and pull out tissue from your kidney which hurts a lot you can just take a urine sample and Probably more accurately predict if your kidney is being rejected or not You can distinguish between certain types of heart failure some which are very lethal and others which are not Just using a quick blood test. Normally that process takes two weeks With using a whole array of methods of testing cardiac function Chronic fatigue syndrome can also be distinguished fairly easily with serum metabolomics EoE we talked about some examples of that in your analysis Packages yesterday with gc autofit and the NMR You can also use it to identify polyps and so here you can look in urine to figure out what's going on in the colon And these metabolites reflect the microbial perturbations that happen in the gut when polyps start developing So in fact, um, a number of companies have started taking Metabolomic markers and translating them into tests Tablon has been working on this Stemina has been working on an autism test and a company in Edmonton has converted those polyp tests that I talked about Into a test that's now used widely in the u.s so Metabolomics is being translated to clinic markers are being discovered and many of these markers have impressive performance sometimes significantly better than gene or protein tests So I think if we can get the message across and if you was Aspiring metabolomic scientists can can make can continue to make interesting discoveries I think there's a good opportunity for labs to adopt quantitative metabolomics for clinical translation I think Better algorithms are being developed using machine learning techniques and some of the methods that Jeff mentioned today that allow you to pick out and pull out more powerful Markers Obviously to make something work in the clinic you want a small number of markers not tens or a dozen You want two or three because that makes the test easy and cheap to perform And as I said a number of them are starting to make their way into laboratory debaunt tests FDA approved tests or european medical agency approved tests And so I think this is an exciting time and and it's really has to be driven by Scientists wanting to see their work translated The last thing before you guys all fall asleep Is to talk about moving metabolomics into drug discovery game So metabolomics began as a drug discovery tool It started in the late 1990s With groups pushing a number of large pharma companies to use metabolomics to assess How Drugs are affecting animals in animal studies Now in drug development it takes about 10 to 15 years to develop a drug So everyone's hoping for a covet 19 drug Unless someone finds a drug that's already being used They are not going to develop a new drug for covet for another 10 years And it's going to cost at least a billion dollars for the company that does it Now it turns out that metabolomics is really useful in drug development It can be used to discover drugs. It can be used to assess the safety of drugs It can be used to assess the response of drugs Or the response of individuals to drugs. It can be used to assess drug testing In phase two and phase three and it can also be used to follow up Patients after the drug has been approved And it turns out that the fat the metabolomics can be used the entire length of the drug development pipeline where Other omics techniques are usually just focused on the discovery side has made metabolomics Useful once again for drug developers And this is just highlighting that the different phases where Metabolomics is is being used It's being used to discover drugs. It's being to help with toxicity screening It's to develop preclinical efficacy biomarkers. It's done to do toxicity assessments in patients It's done to do clinical safety biomarkers and clinical efficacy biomarkers all the way through So no other omics technology in fact no other technology can find utility at every every step of the pipeline Now traditionally the way we do drug discovery is we have used genetics We've used things like GWAS techniques to identify Marker genes many GWAS studies cost millions of dollars and take several years to perform of those Potential genes that seem promising on average about half of them Can be targetable And then of those genes that people find are useful targets only about half of them Can be developed into useful assays involving proteins And then when people start doing high throughput screening to see if they can find a drug Typically only about 20 of those screening assays lead to a lead compound So roughly you've got this 20 success 50 50 percent 20 success just to get to your lead compound And then things get really tough Once you've got your lead compound you have about a one in 500 chance of getting the lead compound through all the different phases of trials And even if you can get the lead compound to complete phase three Even when it's marketed most drugs only succeed about 50 percent of the time Some of them have side effects. They didn't anticipate some of the efficacy isn't as great as they thought And usually someone comes up with a better drug which makes your drug kind of useless So when you look at the total time 20 years the total cost more than a billion dollars and the total success rate Point zero zero one percent if you multiply all those ratios together It's no wonder that most drug companies have stopped doing drug development On the other hand if you use metabolite based drug discovery you have a much better chance So rather than doing GWAS study if you do a metabolomic study to see which metabolites have changed gone up or down Then in fact right away as you guys are learning today. You can zero in on some pathways And you can do that pretty quickly with metabolites. It's not even days. It's more like hours So once you zero it in on the metabolic pathways and therefore some of the critical proteins and genes Then you can start going to databases like drug bank, which lists all of the Known drugs that target certain proteins and many other kinds of databases like Brenda Which allows you to start choosing some potential inhibitors of enzymes or metabolic and metabolic proteins In many cases it turns out those effective inhibitors Probably already exist or have already been discovered and some of them are actually Food-based some of them are nutrient based some of them are already existing as approved drugs And so again using those techniques because we know so much about metabolism And we know so much about how small molecules target various enzymes transporters and proteins It's possible sometimes to get a lead compound quite quickly And what's more is once you've got your lead compound if it's already an approved drug or an already an approved metabolite You can go straight to trials. You don't have to go through the FDA Because it's technically an approved compound And what's more you can use metabolomics to monitor the drug or and its performance Throughout the trials and the tests Now this might seem pretty fancy Fancyful in the sense of yes, could you do this for less than a million dollars? And can you do this in a matter of days or weeks or maybe months? Um Well, in fact, it's actually happened and this was one example which is a discovery made in 2010 roughly about two hydroxyglutarate So it was a metabolomic study They were looking at people who had or developed brain cancers and they found that there is a specific metabolite Two hydroxyglutarate that was elevated in people who had brain tumors And it was also associated with certain people with an inborn metabolism who also developed high levels of cancer But two hg is very abundant in many brain tumors And it turns out that they were able to very quickly figure out what was the enzyme that was causing that and it was a mutant Isosichrite dehydrogenase that leads to the production of this onco metabolite So within a few years literally after they figured this out They went through They did the screens. They found a drug in 2014 Thanks to the metabolomics efforts, and then it was approved in early 2017 They were able to use analogs that had been discovered in the 1970s For studying tc metabolism to model this particular drug and develop it So they're able to use that that large body of history or historical work and known enzyme inhibitors To develop a drug that is already in the market and in fact several other idh inhibitors have already appeared So this is an example of how you can have very accelerated drug discovery where you use metabolomics to Identify the target and use the huge body of metabolism And enzyme inhibition studies that have been published over the last few decades to develop leak compounds very quickly So in terms of the future metabolomics, I think There's several areas that that I think are exciting Areas where a lot of people are working towards where I think is a group and there's new newly trained metabolomics people that you should try and focus on and as I say these are in the areas of automated metabolomics expanding metabolome coverage improving ability to quantify in metabolomics Translating metabolomics not just to the clinic but to other areas from environmental to monitoring to animal Water quality assessments and then trying to get metabolomics Engaged in drug discovery because I think they're clear examples where it's leading to completely novel Targets completely novel drugs and allowing drug development to become much much more rapid than before