Edit 2023/09/06: Many comments so far raise the fact that we can't tell whether there is causation or correlation between an illness and the corresponding microbiome of the patient. My point isn't to solve that problem, correlation on its own is of interest for facilitating diagnostics: If a patient suffering from an un-diagnosed illness X exhibits the a similar biome to that of a patient suffering from a known illness C, then the probability that X = C is higher than if we didn't have that correlation.


The microbiome has been a hot topic for the past decade. This talk gives a good overview of this by highlighting how the rise of "modern" chronic diseases is due to a large part to the imbalances in the microbiome following the rise of antibiotics. The old paradigm of single-pathogen to single-remedy is in need of being replaced by a new paradigm where the root causes of the illnesses lie in entire ecosystems rather single pathogens. This incidentally means that AI and big data are likely the way forward in medical diagnostics.

There is at least a dozen companies (e.g., 1, 2, 3) and research groups (e.g., 1, 2, 3) that are making headway in the field. I understand that a lot of this is still at the research stage and that regulatory hurdles only slow things down. However, from what I gather, most of the actionable outcomes from all this work are:

  • merely descriptive (rather than prescriptive, i.e., they're basically the equivalent of 23andMe, but for the microbiome),
  • tailored probiotics or dietary recommendations that claim to "modulate" the microbiome,
  • therapeutics of Clostridioides difficile. (The few companies that did get FDA approval and financial traction in microbiome analysis ironically all ended up looking a C. difficile.)

I'm probably missing something, but these seem like meager advances compared to the data mining potential of the microbiome. While not dismissing the regulatory and data-sharing challenges, I find it surprising that a systematic data mining of the microbiome hasn't yielded more diagnostic insight in all the aforementioned diseases for which it is known to be relevant.

At the risk of sounding naive, why is microbiome sampling (which is relatively unintrusive) not used systematically by hospitals to profile various illnesses? Such data would be a gold mine of insight as opposed to the vary basic tests commonly done (e.g., blood tests or imaging). (Even fecal sampling, to my knowledge, is very rarely used.)

  • 1
    Part of the problem is that its only been in the past 10-15 years that microbiome analysis has been easy and cheap, and it still costs in the $100/sample range, even if you have a sequencer . Another part of it is that this sort of analysis hasn't been validated for clinical use. You might find this paper of some interest.
    – bob1
    Sep 6, 2023 at 21:04
  • 1
    re: your edit - what are the cases you know about where there is evidence that the microbiome specifically provides additional diagnostic information, above and beyond what is typically used? Not just where it is associated with something, but where it is shown useful in diagnosis.
    – Bryan Krause
    Sep 7, 2023 at 15:57
  • 1
    re: your edit - You've made an erroneous statement/assumption. You're not medically trained or qualified to state that correlation is important (or "interesting") for diagnosing illnesses. Just because you wish it to be so does not mean it is true. Sep 7, 2023 at 21:38
  • 2
    @Tfovid - Don't be so patronizing. I've practiced medicine for decades. You watched a video. Until we have more evidence (not correlation), that is all we have: correlation. Maybe even coincidence. Maybe something else entirely is affecting both the body and the microbiome. You have nothing here to tie your high horse to. You're not correct because you wish to be. Sep 8, 2023 at 8:08
  • 2
    @Tfovid Why would you need a microbiome to diagnose obesity when a scale works?
    – Bryan Krause
    Sep 8, 2023 at 13:06

1 Answer 1


I've been following microbiomes since they first began to be discussed, and remember the first fecal transplant done for treatment of C. diff. Now it's a common enough treatment. But when an average weight woman with C.diff had a transfer from her daughter, who was obese, became obese, that was really astonishing (this was in 2015). By then, many researchers in the US and the EU were doing microbiome studies.

The NIH funded Human Microbiome Project (HMP) Consortium has established a population-scale framework which catalyzed significant development of metagenomic protocols resulting in a broad range of quality-controlled resources and data including standardized methods for creating, processing and interpreting distinct types of high-throughput metagenomic data available to the scientific community. Here we present resources from a population of 242 healthy adults sampled at 15 to 18 body sites up to three times, which to date, have generated 5,177 microbial taxonomic profiles from 16S rRNA genes and over 3.5 Tb of metagenomic sequence.

That was in 2012. The NIH is still funding the HMP.

I'm not how sure how quickly you think research can/should be done, but so far, there has been a correllation between gut microbiota and intestinal bowel diseases, coeliac disease, irritable bowel syndrome, colorectal cancer, chronic liver diseases, pancreatic disorders/cancer, liver cirrhosis, obesity, type 2 diabetes and non-alcoholic fatty liver disease among others.

But which came first, the chicken or the egg? We know the answer for Helicobacter pylori and obesity, but not for the rest. That would require, among other things, longitudinal studies starting at birth, which would involve unimaginable expense.

...why is microbiome sampling (which is relatively unintrusive) not used systematically by hospitals to profile various illnesses?

The gut alone is populated by from 10^7 to 10^14 microorganisms. Hospitals do not have money coming out the wazoo, nor are they equipped to do that kind of analysis. For that matter (money), neither do most governments, which is why many countries have national debts. Money is needed for those studies, and it must come from a limited budget. Increased spending in one area means decreased spending in another, or more borrowing.

This incidentally means that AI and big data are likely the way forward in medical diagnostics.

Please know that as little as I know about AI, I do know that it is abysmal when it comes to health/medicine. IBM spent many billions on Watson (unintentionally very well named) with the goal of improving health care. It's been pretty much a disaster, with Watson scoring at or below the proficiency of second year medical students, beating them in two areas only: diagnosis of breast and (I think) lung cancer, but never reaching the proficiency of the lowly first year resident. MD Anderson spent $95M on Watson but trashed it because it just wasn't adding anything to health care except more work and headaches. But I am not fluent in AI, so, hey, who knows?

A framework for human microbiome research

  • That last paragraph is actually what astonishes me here. I don't know what kind of blunders they made with Watson, but coming from a data science background, the fact that the microbiome isn't systematically looked at by modern AI---given the fact that it is a big data problem by definition---seems like a major oversight to me. AI now isn't the AI from 2012.
    – Tfovid
    Sep 3, 2023 at 15:39
  • 3
    Great answer. Agree with the point about correlation. AFAIK we have only correlated changes in microbiome with disease; causation (or whether the microbiome is altered because of disease rather than the other way around) remains to be demonstrated, reducing usefulness for diagnosis.
    – Chris
    Sep 3, 2023 at 16:07
  • 2
    @Tfovid - Do you think IBM, having spent 200 billion dollars on Watson so far, haven't used data scientists? AI is just that: artificial. They sold Watson Health at a great loss. I'm sure they tried their best. But as I said, I am not very familiar with AI. AI in languages is great. In medicine? not so much, even in 2023. Sep 3, 2023 at 16:43
  • 2
    @Tfovid AI doesn't care for Occam's Razor, and doesn't know when to do a controlled trial. I get why you're excited, but your description of what it would do hits the nail on the head: correlation does not imply causation. (AI today is basically AI from 2012, outside a few narrow areas.) This would be useful for helping find research hypotheses, but probably not useful as a diagnostic. (Or maybe it would be: who knows?)
    – wizzwizz4
    Sep 3, 2023 at 19:34
  • 1
    Capabilities of 'AI' isn't the only problem. Even if you ignore all the problems with actual analysis, and assume you have a magic box, you still have a huge logistical/engineering challenge performing the capture of the microbiome in a way that generates digital data that can be processed by AI system to perform diagnostics. There's a reason the deployment of 'AI' systems in healthcare has been focused on operational and radiology workflows. These are the fields that already work with primarily digital data. Sep 3, 2023 at 20:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.