Generative AI for Analysis of Gut Microbiome Tests and Personalised Treatment Advice

[SpacePenguin]

Can you remember for how many years you've been on such intakes, before your test in 2018?

I was consuming the sencha/choc/berries for about 10 years before the test. Before the sencha, I was drinking gunpowder green tea for 4 years before that. Gunpowder has less EGCG.

 

[Violeta]

I wonder if Vitamin D would help.

Interesting about prausnitzii: Faecalibacterium prausnitzii (F. prausnitzii) is a gut bacteria that may play a role in vitamin D production and metabolism.

And: Lactobacillus strains, particularly Lactobacillus rhamnosus GG, can positively influence vitamin D absorption and levels in the body.

Vitamin D supplementation can increase Roseburia abundance: Studies show that Vitamin D supplementation can lead to an increase in the abundance of beneficial bacteria like Roseburia in the gut.

I first saw this study about it promoting increase in Bacteroidetes,
Specifically, the Bacteroidetes to Firmicutes ratio increased, along with the abundance of the health-promoting probiotic taxa Akkermansia and Bifidobacterium.

So the increase in Akkermansia might not be wanted but the increase in Bifidobacterium and Roseburia would be beneficial.

Too bad there isn't a product for F. prausnitzii.


https://www.nature.com/articles/s41598-020-77806-4

The potential role of vitamin D supplementation as a gut microbiota modifier in healthy individuals​

In summary, while the exact mechanisms are still being explored, research indicates a link between Eggerthella lenta and vitamin D. Specifically, the absence of the vitamin D receptor is associated with increased Eggerthella in the gut. This relationship likely stems from vitamin D's influence on the overall gut microbiome composition and integrity.

 

[SpacePenguin]

I wonder if Vitamin D would help.
Too bad there isn't a product for F. prausnitzii.

I've taking D since before onset of CFS and my D status has never been a problem. There are literally thousands of interventions one might consider that should have some beneficial effect on one's specific gut bacteria. For each intervention, you really need to consider how it will affect the ecosystem as a whole, which means assessing impacts on potentially thousands of species (plus other effects). Sorting through that massive amount of data is precisely the problem ChatGPT, etc can address.

Ah, yes, the wonderful but fragile F. prausnitzii. I have been focused on doing all I could for them for years. But it seems that anything I was doing was have been thwarted by the support I was giving he Akkermansia and Flavonifractor. I'm not holding out hopes for an f.prau probiotic anytime soon - it is just too sensitive to oxygen. Faecal microbiota transplant seems to be the only viable way to introduce it. I just hope it is still present in my gut, but just below the detection threshold of the test.

 

[pamojja]

The species list is very extensive. 60 were detected in my sample, but I had lots that were tested for but undetectable

Considering that western man contains up to about 1000 spezie

For each intervention, you really need to consider how it will affect the ecosystem as a whole, which means assessing impacts on potentially thousands of species (plus other effects). Sorting through that massive amount of data is precisely the problem ChatGPT, etc can address.

That is exactly the main problem I wanted to point out. If a microbiome test lists only 60 spezie out of probably up to 1000, ChatGPT is not able to sort out interactions to the majority, especially with the unknown.

One can experimentally try to alter a few of the known 60 spezie at a time only. Which might work out to health-improvements, or not. But backfire with other specie, known and unknown.

It's still sort of gambling, since ChatGPT too can't predict interactions to the unknown.

However, gambling with experiments is the only thing one can do, if all other interventions didn't help.

 

[almost]

My testing was by Microba (now called Co-Biome) in Australia in 2018. The species list is very extensive. 60 were detected in my sample, but I had lots that were tested for but undetectable and very low diversity.

Hi @SpacePenguin, thanks for the interesting thread. I've followed it with some interest, as I'm doing something similar. Sorry that you're in the position to have to do this, however.

First, I generally support what you're doing. Yes, there are risks, but you know them. My concern is the age of the data. If you are making decisions based off the 2018 test, that data is of questionable use, unless you have kept the exact diet, lifestyle, and general health since then. You can change your biome significantly in just a couple of weeks with diet. I would suggest a retest with a company that gives you the raw data. Better data would lead to better decisions.

The other thing I have found is to ask it LOTS of questions around your core question -- have a conversation. Ask about the general dearth of species, ask about gut ph levels, would more acid or alkaline help (more acidic usually favors the ones we think are beneficial), would certain herbs help, would doses of Lactobacillus like one gets with Dr. Davis' yogurts set a good stage, etc. They can be transient, but can favorably reshape the landscape.

I too doubt that the green tea, berries and chocolate alone led to the overgrowth of Akkermansia (I did the same, and my level is 32nd percentile). If you only have 60 identifiable species, then there isn't much to push back against the Akkermansia, for example.

Thank you again, and good luck.

 

[SpacePenguin]

My concern is the age of the data. If you are making decisions based off the 2018 test, that data is of questionable use, unless you have kept the exact diet, lifestyle, and general health since then. You can change your biome significantly in just a couple of weeks with diet. I would suggest a retest with a company that gives you the raw data. Better data would lead to better decisions.

Thanks, for your thoughtful response, almost! Yes, of course, the age of the data is not ideal. But the point of my experiment was partly to see how much mileage I could get out of the data I had since the same test now costs AUD 485 (~315 USD). In any case my diet, lifestyle, bowel movements, and health are much the same now as in 2018 when the test was done, so I suspect that my gut might not have changed much. In any case, my plan was to try ChatGPT's suggestions and then retest in 6 months. I couldn't afford to do that test more than once a year, and even that is painful. Ken Lassessen recommended I try Biomesight's testing since they have a code for ME/CFS ("mecfs") which brings the cost down to 95 USD, and they take international orders. So I have a kit on the way now and at that price I could do it at 6 month intervals. Biosight's testing methodology (16s) is not as good as Co-Biome/Microba's (shotgun metagenomics), but it is still good enough apparently and better value.

The other thing I have found is to ask it LOTS of questions around your core question -- have a conversation.

Yes! This is the great thing - you can really drill down into details and ask it about how general research relates to your own situation. You can ask it how strong the evidence is for its claims and what it's specific sources are.

would doses of Lactobacillus like one gets with Dr. Davis' yogurts set a good stage, etc. They can be transient, but can favorably reshape the landscape.

NOOO! Here again is a point where ChatGPT's ability to draw upon and synthesise a huge amount of data is very advantageous. The doctors who had advised me on this (who have a very good knowledge of CFS and gut stuff and who I trust a lot) had advised me to just try lots of different probiotics, cycle through them, and just stick with one if it proves particularly beneficial. But ChatGPT tells me that most "Lactobacillus" (I realise the term is antiquated now) actually promote Akkermansia and are mucolytic (they break down mucus, my gut mucus is presumably thin and my mucosa fragile given the composition of my microbiome). So Lactos would actually be damaging until I can improve things a bit. Instead, ChatGPT's advice is that I should focus on Bifidobacteria at first. The problem is that most probiotic formulas emphasise Lactos. I asked it why this is, and it said it is purely because Bifidos are harder to culture and thus more expensive and not at all because Bifidos are more potent or make up a smaller portion of the microbiome.

 

[SpacePenguin]

That is exactly the main problem I wanted to point out. If a microbiome test lists only 60 spezie out of probably up to 1000,

Microba/Co-Biome testing is shotgun metagenomics, which is the best you can get. On their site, they claim their test "Can report on over 28,000 species although a typical healthy sample will contain between 110 – 244 species." The fact that I only have 60 should make it simpler to work out what to do.

Worth keeping in mind, that there will be species present that do not show up on the report because their numbers are below the detection threshold. So hopefully I still have some f.prau etc knocking around in there.

 

[SpacePenguin]

It's still sort of gambling, since ChatGPT too can't predict interactions to the unknown.
However, gambling with experiments is the only thing one can do, if all other interventions didn't help.

The analogy to gambling is overdrawn in my view. In actual gambling you stand to lose as much as you stand to gain (in a fair bet, at least). You stake 1 dollar on the roulette wheel which you might lose, or you might win $2. There is no such jeopardy involved in using any of the things that have been recommended to me: PHGG, lactulose, slippery elm, psyllium husks, glutamine, bifidobacteria probiotics, lactoferrin, nattkinase & serrapeptase. All very innocuous - would probably be beneficial for most people. None of these things has the remotest chance of worsening my condition. While I am spending money, I wouldn't consider it "losing money" since there is an expected benefit.

 

[pamojja]

Microba/Co-Biome testing is shotgun metagenomics, which is the best you can get. On their site, they claim their test "Can report on over 28,000 species although a typical healthy sample will contain between 110 – 244 species." The fact that I only have 60 should make it simpler to work out what to do.

Your specific test seems able to find even 3 times more species, than what @Lassesen considers the most extensive. (https://precisionbiome.eu/ deep shotgun metagenomics, with up to 8,000 microbes, incl. bacteria, fungi and virus).

Here one can see, that Microba had only 28 samples uploaded, 1 sample with the highest of 207 specie only.

I calculated from the added up percentages of my 74 found specie, that it identified 44% percent of all, therefore 56% unknown. In my case (phylum and class was identified 100% in my case, order still 90%, family 88%, and genius 77%)

What is your percentage added up from all % of your specie?

Worth keeping in mind, that there will be species present that do not show up on the report because their numbers are below the detection threshold. So hopefully I still have some f.prau etc knocking around in there.

Exactly that's why I calculated from the data, how much % are unknown, in my test.

The analogy to gambling is overdrawn in my view. In actual gambling you stand to lose as much as you stand to gain (in a fair bet, at least).

Well, I use the gambling analogy, due to bad health (bodily disability and loss of independence) as "nothing more to lose". With nothing more to lose, one can still win in gambling

Losing money is nothing, compared to health.

 

[pamojja]

Considering that western man contains up to about 1000 spezie, already mingling Hadza Hunter-Gatherer up to 2000, and formerly uncontacted Yanomani tribe up to 4000 distinct bacteria.

Checked for sources, and found this claim unfounded. Just imprecise reporting. Sorry for not checking first.

Averages for Western populations and Hadza are well supported by research, while such absolute numbers are not available for the Yanomami in the same way.
Here’s a clear summary of the evidence:

• Western Populations:
  • Average species per individual: Multiple studies report that Westerners typically have between 150 and 400 bacterial species per person, with some ultra-deep sequencing studies in healthy adults finding averages around 193–317 species per individual56.
  • Example: One large study found an average of 193 species per person (standard deviation 55.1), with a range from 58 to 3465.
  • Other studies: Another found an average of 186 ± 51 species per person6.
• Hadza Hunter-Gatherers:
  • Average species per individual: Recent studies using deep sequencing report averages around 730–800 species per person in the Hadza, which is significantly higher than in Western populations56.
    • Note: Some sources may use operational taxonomic units (OTUs) or species-level phylotypes (SLPs), but the Hadza consistently show much higher diversity than Westerners.
• Yanomami:
  • No absolute average per individual: While the Yanomami are shown to have the highest microbiome diversity ever reported in humans, peer-reviewed studies do not provide a specific average number of bacterial species per individual, only relative diversity metrics and comparisons56.
  • Relative, not absolute, diversity: The landmark study emphasizes that Yanomami have “unprecedented levels of bacterial diversity,” but does not give a figure like “X species per person.”

In summary:
Averages for Westerners and Hadza are supported by published studies and are commonly reported. For the Yanomami, only relative diversity is reported, not absolute species counts per individual, despite their microbiome being recognized as the most diverse in humans56.

1. https://pmc.ncbi.nlm.nih.gov/articles/PMC8497558/
2. https://www.nature.com/articles/s41467-022-31502-1
3. https://www.sciencedirect.com/science/article/pii/S2001037023004622
4. https://pmc.ncbi.nlm.nih.gov/articles/PMC8749183/
5. https://www.biorxiv.org/content/biorxiv/early/2019/02/27/557124.full.pdf
6. https://www.frontiersin.org/journals/microbiology/articles/10.3389/fmicb.2020.02029/full
7. https://www.sciencedirect.com/science/article/pii/S2212877816301004
8. https://www.nature.com/articles/s41392-024-01946-6

 

[pamojja]

"Can report on over 28,000 species

Now, I also understand why anyone can make such claims. Specie is the common parlance for bacteria, but actually reported is something else:

Spoiler: OTU(ASV vs. specie

Your question touches on a crucial distinction in microbiome science: the difference between microbial species as traditionally defined (organisms with unique physiology and ecology) and operational taxonomic units (OTUs) or amplicon sequence variants (ASVs), which are defined purely by DNA sequence similarity124.

OTU/ASV vs. Species: What’s the Difference?​

• OTU (Operational Taxonomic Unit):
  • Definition: Groups of sequences (usually 16S rRNA gene fragments) that are at least 97% similar to each other.
  • Interpretation: Each OTU is intended to approximate a bacterial species or genus, but in practice, one OTU can lump together several real species or strains, or even unrelated sequences, depending on the clustering threshold14.
  • Used in: Older and some current microbiome studies, especially when computational power or reference databases are limited.
• ASV (Amplicon Sequence Variant):
  • Definition: Unique sequences that differ by even a single nucleotide; each ASV represents a distinct sequence variant.
  • Interpretation: ASVs are more precise than OTUs, capturing subtle genetic differences that may correspond to different species or strains, but not necessarily to named, biologically defined species124.
  • Used in: Modern microbiome studies for higher resolution and reproducibility.
• Species (Traditional Biological Definition):
  • Definition: A group of organisms that can interbreed (for eukaryotes) or, for bacteria, a group with similar genetics, physiology, and ecology.
  • Interpretation: In microbiome research, “species” are often only inferred from DNA sequences, and many sequences do not match any named species in databases.
  • Used in: Formal taxonomy and sometimes in microbiome reports, but rarely as the primary unit in research due to limited reference data.

How This Affects Your Perception​

• OTU/ASV Counts Are Not the Same as Named Species:
  • When you see “species” counts in microbiome research, these are usually OTUs or ASVs, not formally named species. This means the numbers are higher than the actual number of named, biologically defined species present124.
  • For example, a Westerner might have 200–400 OTUs/ASVs, but only a fraction of these correspond to named species. Your personal report with 91 named species is more restrictive but arguably more meaningful in a traditional biological sense.
• Extinction Event Analogy:
  • The idea of an “extinction event” in the Western microbiome is based on the loss of diversity measured by OTUs/ASVs, not necessarily the loss of named species. However, since many of these OTUs/ASVs may represent unique strains or even sequencing noise, the real biological impact is hard to quantify12.
• Normal Perception vs. Scientific Reality:
  • The general public often thinks of “species” as distinct life-forms, but in microbiome research, “species” are usually just unique DNA sequences. This can lead to confusion when comparing research results to personal microbiome reports or traditional biological concepts.

Summary Table​

Concept	Definition	Example in Microbiome Research
OTU	Cluster of similar sequences (e.g., 97% alike)	200–400 per Westerner (not species)
ASV	Unique sequence variant (single-base difference)	200–400 per Westerner (not species)
Species	Biologically defined, named organism	~91 named species in your report

Key Takeaway​

OTUs and ASVs are not the same as named species. They are proxies used in research to estimate diversity, but they do not always correspond to distinct life-forms. This is why your personal microbiome report (using named species) gives much lower numbers than research averages (using OTUs/ASVs). The “extinction” of microbiome diversity in Westerners is real in terms of sequence diversity, but it is not the same as the loss of named, biologically defined species124.

1. https://www.zymoresearch.com/blogs/blog/microbiome-informatics-otu-vs-asv
2. https://pmc.ncbi.nlm.nih.gov/articles/PMC8386465/
3. azI9taClDhQ
4. https://www.cd-genomics.com/resource-amplicon-sequencing-analysis-out-asv.html
5. https://www.frontiersin.org/journals/microbiology/articles/10.3389/fmicb.2022.796758/full
6. https://pmc.ncbi.nlm.nih.gov/articles/PMC9029325/
7. https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0309065
8. https://www.mdpi.com/2073-4425/12/4/564

So only defined bacteria's sequences show up in test results. And my erroneous numbers in different populations derived from all bacteria defined and found - not from individuals - but distinct bacteria found in whole populations.

Spoiler: relationship between

There are no widely accepted, direct conversion factors between OTUs/ASVs and named bacterial species in microbiome science. However, there are some general relationships and caveats that help explain where numbers like 1,000, 2,000, or 4,000 “species” in popular reporting might originate.

Relationship Between OTUs, ASVs, and Species​

• OTUs (Operational Taxonomic Units):
  • Definition: Groups of sequences with at least 97% similarity (sometimes 99%), intended to roughly approximate a bacterial species or genus246.
  • Reality: One OTU can include multiple real species or strains, or even unrelated sequences, depending on clustering thresholds and reference databases24.
• ASVs (Amplicon Sequence Variants):
  • Definition: Unique sequences that differ by even a single nucleotide.
  • Reality: ASVs are more precise than OTUs, but each ASV still does not necessarily correspond to a named species—many ASVs are not assigned to any known species45.
• Species (Named):
  • Definition: Formally described bacterial species with unique physiological and ecological traits.
  • Reality: Only a fraction of OTUs or ASVs can be matched to named species, especially for poorly studied environments or non-model organisms46.

Estimates and Conversion​

• No Direct Conversion:
  • There is no reliable formula to convert OTUs or ASVs to named species counts.
  • In well-studied environments (like the human gut), a substantial proportion of OTUs/ASVs can be assigned to genus or family, but far fewer to named species.
• Typical Ratios:
  • OTUs: Often overestimate diversity compared to named species.
  • ASVs: Can be even more numerous than OTUs in some analyses, but tend to better reflect true genetic diversity without lumping24.
  • Named Species: Far fewer than OTUs or ASVs in most samples.

Where Do the Big Numbers Come From?​

• Community-Level (Gamma) Diversity:
  • The numbers 1,000, 2,000, or 4,000 “species” sometimes cited in popular articles refer to the total number of OTUs or ASVs found across an entire population or community—not per individual5.
  • For example, if you pool all samples from a group (like the Yanomami), you might reach thousands of unique OTUs or ASVs, but each individual will have only a fraction of these.
• Misinterpretation in Media:
  • Popular reporting sometimes conflates community-level diversity (total OTUs/ASVs across many people) with individual diversity (OTUs/ASVs per person).
  • This can lead to exaggerated claims about “species” per individual.

Summary Table​

Metric	Typical Count (Human Gut)	What It Represents	Notes
OTUs	200–400 per individual	Clusters of similar sequences	Overestimates named species
ASVs	200–400+ per individual	Unique sequence variants	More precise, but still not species
Named Species	~100–300 per individual	Formally described bacteria	Only a fraction of OTUs/ASVs
Community OTUs	1,000–4,000+ (total)	Across all samples in a group	Not per individual

Spoiler: key takeaway

There is no direct conversion between OTUs/ASVs and named species. The large numbers (1,000, 2,000, 4,000) you see in popular reporting likely refer to total community-level diversity across many people, not per individual. Each person typically has hundreds of OTUs/ASVs, but only a fraction of these correspond to named species. This is why your personal report (with named species) is much lower than the “species” counts you might read about in the media245.

1. https://pmc.ncbi.nlm.nih.gov/articles/PMC9029325/
2. https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0309065
3. https://www.mdpi.com/2073-4425/12/4/564
4. https://www.zymoresearch.com/blogs/blog/microbiome-informatics-otu-vs-asv
5. https://www.frontiersin.org/journals/microbiology/articles/10.3389/fmicb.2022.796758/full
6. https://pmc.ncbi.nlm.nih.gov/articles/PMC8070570/
7. https://ojs.library.ubc.ca/index.php/UJEMI/article/download/197219/192176/244048
8. https://pmc.ncbi.nlm.nih.gov/articles/PMC11449282/

 

[almost]

NOOO! Here again is a point where ChatGPT's ability to draw upon and synthesise a huge amount of data is very advantageous.

I hope you understand I wasn't making a specific recommendation for you. Just throwing out ideas from different ends of the spectrum, not specific to you, to provide scope in questions to ask. Sounds like you've done that. Of course, you get to decide on your own course of action.

I do understand that Lactobacillus can be a problem. The ubiquitousness of L. acidophilous, for example, is a plague on the probiotic world, as it is cheap but can really muck things up in a sick person. But not all lactos are equal. There are some that help establish Bifido for example.

I do wish you well.

 

[pamojja]

I do wish you well.

I hope it is understood that most of us do. I do, of course, wish that your experiment you decided for, do work for you!

And different perspectives on issues is no expression of ill-will at all.

 

[Violeta]

When one has thin mucus lining in the gut, looking at something for glycocalyx or glycosylation may be helpful.

Most possibly helpful thing I've found so far is fucoidan. Although it might be helpful to look at sialic acid, too.

Do you know what causes thin mucin lining? I am seeing that glycosylation issues can be involved.

@mariovitali has some information on Twitter (X) about glycosylation as it is related to ME/CFS.

 

[Wishful]

None of these things has the remotest chance of worsening my condition.

I wouldn't be too sure about that. Since we don't know what is involved in ME, anything has a potential to be beneficial or detrimental. Something beneficial in one way, such as helping a normally "good" strain might end up reducing another strain that is actually beneficial wrt your ME. No matter how safe something is theoretically supposed to be, if it makes you feel worse, pay attention to that.

 

[Wishful]

When I wanted to try some probiotics, I was quite disappointed at the limited offerings. Where are the entrepreneurs offering strains that aren't in the common brands?

I wonder how many human-compatible strains exist in edible insects. Those hard exoskeletons might act as time-release capsules if you need to get those strains down to your colon. Chew for upper-intestinal tract species.

None of the flying insects I accidentally swallowed helped.

 

[pamojja]

None of the flying insects I accidentally swallowed helped.

Wow, what an experimenter. Never thought about it, but our ancestors definitely ate insects too. However, they also allegedly ate up to 200 different plants. A substrate for most varied bacteria, not available to us today.

I do understand that Lactobacillus can be a problem.

With none in my biome (compared to fermented food eaters, as I actually always have been), I do benefit from high count lacto- and bifido- brands. Though, I only take them after night shifts on a empty stomach before a further nap. And off I go, into grotesque continuous and at times lucid dreaming land. Without, I don't remember dreams. Though B6 and resistant potato starch were first needed, for this effect to occur now.

 

[LINE]

Barrier integrity is a foundational must - why? Without barrier integrity, inflammatory cytokines are highly expressed. This is aka leaky gut. The barrier becomes eroded due to immunity - e.g. you contracted a virus, bad bacteria and the like. The immune system is busy trying to control the event - byproducts of immune activation result in barrier integrity dwindling.

With the high oxidative stress, the SCFA and bifido will not populate correctly. SCFAs produce butyrate naturally which keeps the barrier integrity strong but this is a bad loop b/c SCFAs cannot populate the butyrate b/c of the barrier integrity leakage.

Bone broth contains glycine, proline and threonine which helps rebuild the integrity as does anti-inflammatory polyphenols (which you are doing). The broth (or stock) is the same protein complex as the gut (collagen based). You can also look at collagen peptides but for my money, the bone broth (or stock) was better.

Antioxidants will also protect the barrier; Vitamin C is a good bet (I would suggest Quali C which is superior IMO and competitively priced. There is some current research on creatine @10g per day, there is a loading phase of about 10 days to notice changes.

Stay on the milk thistle to keep the liver open - the liver takes a beating from all of the oxidative stress.

 

[SpacePenguin]

I hope you understand I wasn't making a specific recommendation for you.......I do wish you well.

Oh yeah, no worries! I just had a visceral reaction of pain at the idea of someone making the same mistakes I seem to have made (even though it was for all the right reasons).

 

[SpacePenguin]

When one has thin mucus lining in the gut, looking at something for glycocalyx or glycosylation may be helpful.

Most possibly helpful thing I've found so far is fucoidan. Although it might be helpful to look at sialic acid, too.

Do you know what causes thin mucin lining? I am seeing that glycosylation issues can be involved.

@mariovitali has some information on Twitter (X) about glycosylation as it is related to ME/CFS.

I have seen those things mentioned over the years, but always thought the evidence was thin and the cost was high. There's so many things one might try, and I have done over a 100 interventions without much luck.