My uBiome Gut Test Results: How Well Does My Gut Function?

Out of the six uBiome gut samples I took before, during, and after a 12-day Elixa Probiotics ( regimen (posts here, here, and here) I’ve received the results on five of them. One sample was damaged in shipment (the final test, 2 weeks from completing the course) and uBiome has sent a replacement. Still waiting on results from American Gut (one of which was taken from exactly the same poop sample as the uBiome kit that was damaged).

We’re still looking to crunch raw data and taxonomy, but I can pretty much tell it’s about what I always thought of these tests. They really don’t tell you much useful, they don’t seem to correlate with anything concrete I can think of, and they’re all over the map. For instance, in my first test, my “diversity” was in the 80th percentile. Jump for joy, right? Well, then 10 days later, the next test (and this was right before even starting Elixa), and diversity is 12th percentile.

Then, during the Elixa 12-day course, in the first week, massive shifts in bacteriodetes and firmicutes, The former went from 30 to 80% of my microbiome and the latter, correspondingly, from 66 to 16%, so basically a complete reversal in the space of a week. Elixa? Well, perhaps, but then another week, i.e., at completion of the course, firmicutes were up to 30%. If Elixa were the cause of reducing them so dramatically after 1 week, then seems like they should have at least stayed there the 2nd week, not basically doubled.

This is all without further analysis (just realized I had results yesterday) or with digging into it with Karl at Elixa, or anyone else. I’m trying to get stuff into MG-Rast, but it’s a government funded deal, so it’s like working with websites designed in 1995.

Getting the AmGut results are likely to shed even more confusion on the whole deal, since I believe it was Tim Steele who submitted swabs to both uBiome and AmGut from the same TP swipe, and got significantly different results. But this had to be done and who knows, perhaps there will be earth shattering information, but I doubt it. Now, at least I’ll be able to say that I did the tests, showed the results. But then I’ll add that your diet is probably the most profound influencer, and the Elixa probiotics might be useful and if they help you with some problem: make you feel better, sleep better, poop better, or have better skin, then use them from time to time. Of course, there are many self-reported anecdotes in the various posts attesting to exactly those effects, with damn few reporting negative outcomes, or no effects. Perhaps the best indicator is the significant number of re-orders Karl is getting.

…I’ll suggest that this graph might actually be the best indicator of overall gut health for me. It’s a Function chart based on the sequenced bacteria, as some species are better at some things than others. Click on the image for the larger size.

Screen Shot 2015-11-09 at 4.16.41 PM

What that looks like to me is a pretty functional gut for a wide variety of things like carb metabolism, amino acid metabolism, etc. And when compared to the general population, no big surprises except that I’m apparently more functional than average at carbohydrate metabolism and less functional in fatty acid metabolism.

I think at this point I’ll hold off on taking the uBiome replacement test until I do something remarkable, like perhaps the potato diet for a week or two. If the test is way different from the general baseline I’ve established, might tell us something.

More later, should digging into the analysis bring up anything noteworthy.

My main reason for posting this is for a note of caution in the bright-eyedness of this stuff. I fell into it myself, for a while. But when you’re dealing with something like 3 million genes, or whatever it is, 500-1,000 species in 100 trillion bugs, there just might be enough diversity to claim anything you want. It just comes down to finding and emphasizing that one species that’s your magic species or demon species, or whatever, and you’ve made your case.

You won’t get that here. How you feel, sleep, digest food and eliminate are the solid results you want to focus on.

Richard Nikoley

I'm Richard Nikoley. Free The Animal began in 2003 and as of 2021, contains 5,000 posts. I blog what I wish...from health, diet, and food to travel and lifestyle; to politics, social antagonism, expat-living location and time independent—while you sleep—income. I celebrate the audacity and hubris to live by your own exclusive authority and take your own chances. Read More


  1. Jon on November 10, 2015 at 02:10

    To me there’s four useful things you can learn if you do multiple tests:

    1. Are you consistently lacking certain specifies of beneficial bacteria across all tests
    2. Are you consistently showing reasonably high levels of bacteria considered potentially problematic
    3. Are you effecting a change over time into a more beneficial profile (or what we believe at the moment is beneficial)
    4. How you compare to other groups.

    These are only really needed by someone trying to resolve a gut related problem. Put together with other types of tests, they may provide a more complete picture of the underlying cause behind problems.

    In my own case, my uBiome showed an incredibly small number of Bifido, Lactobacilli and no beneficial E.Coli. I’m trying to resolve a number of symptoms including mild fatigue. My missing/low species match some of the reported profiles of CFS. I’m supplementing with E. Coli Nissle strain, high bifido, me lacto and GOS. So far so good, I’m improving. Would I have tried the current protocol without having done the uBiome? It’s unlikely. The bifido/lacto yes, but not the e.coli.

    Having said that, it does seem our knowledge is very immature at this point.

    • Natalya on November 19, 2015 at 16:26

      Hi Jon,
      Are you using the German product, Mutaflor (E.coli Nissle 1917)? I made yogurt w it for some months, made very well on stove top w pilot lights. When I had a Biomatrix parasite test done they remarked that I had high concentrations of E. coli, I figure it is the yogurt I made.

    • Jon on November 19, 2015 at 21:13

      That’s right, I’m using Mutaflor. Good on you for making yogurt with it. If you had suggested E.coli yogurt to me a month ago, would have thought you were mad, but I think the Nissle 1917 strain is a good probiotic. It certainly has a bunch of studies to back it.

  2. Karl S on November 10, 2015 at 04:48

    I like the writeup, Richard. And I think it was completely worth a shot anyway!
    I agree 100% with your conclusion – qualitative, anecdotal results may have to suffice for now. (In addition to other tests which measure stool for metabolic byproducts, pH, and other variables)

    *Preface: The following is only relevant based on the assumption that there is a flaw in the test procedure. There may be none. I don’t want to insult anyones hard work or condescend a company that knows way more about microbiome assaying that myself! These are just my thoughts on the matter.*

    It may be the case that a microbiome assay would become closer to being representative only if the participants were to take 1-3 days worth of poop, blend it all up until it is completely homogeneous (in terms of bacterial distribution) and then take a sample from that. Caveat being: not a procedure that a consumer is likely to perform.

    uBiome has raised several million dollars of investment capital. Comparatively speaking, the sales of the kits alone can’t amount to much. I gather that their plan is to generate revenue via partnerships with other companies who want access to their data. (I may be mistaken here.)

    So, for example, Procter and Gamble come along and pay money to be able to see which microbial profiles are correlated with which conditions/health markers. Then they go about developing methods to culture the key bacteria responsible for certain health markers, or developing targeted antimicrobials that kill only the bacteria responsible for certain conditions.
    Unfortunately this may be a hard sell if the data is not representative of the entire colonic microbiota. If there is so much variation from sample to sample, it is likely due to the thoughts I mentioned in a comment a few weeks ago:
    ‘Because another unknown (in addition to the assumption detailed in Richard’s subsequent post) is the degree of homogeneity of the microbiotic makeup of a stool sample. Or more simply: How well mixed up are the microbes within a poop (?). Will a small sample represent the sample as a whole (?). Will 3-4 swab sites represent the sample as a whole (?).

    Are we behaving with the naivety of an alien craft that has made a quick pit-stop on earth, excavated only 1 cubic mile, and from there tried to characterise the entire ecosystem of the planet? If that is so, then a cubic mile of NYC versus a cubic mile of Queen Elizabeth National Park Uganda is a good metaphor for how differently the apparent ecosystem of Earth would be characterised!

    So it’s not that AmGut or uBiome would have inaccurate tests. In fact I am confident they are among the best two institutes for this type of testing on the planet! It just might be the case that the concept of profiling an entire stool from a small swab sample does not work out in practice.’

    No problem is without a solution, however 🙂
    I’ve had a think about it and – weighing up logistics, cost, practicality/appeal-factor – a simple tweak to the collection technique may be all that is needed.
    It could go something like this:
    Mail out 1-3 suitable containers and 1 return mailer box.
    Require each customer to collect an ENTIRE stool sample (even better would be one per day for 3 days) and seal it in the suitable container. The critical difference here is that it is an ENTIRE stool sample, not a smear.
    Each plastic container would be sealed and could be rinsed on the outside and put somewhere for the next 1-3 days. The seal would block odour egress of course. These containers would have some liquid to cease all microbial proliferation from the moment the stool enters the container.
    When all samples are collected – let’s say 3 – put them all in the pre-paid mailer box and post it back.
    That way all the blending would be done at the company rather than the consumer having to do it. The blending would be the main detractor for consumer appeal, so this circumvents that concern.
    Now when the samples arrive back at uBiome HQ they would simply blend all contents of all containers together, until homogeneous. A lovely job for a new intern perhaps 😉 Then they could take a suitable volume of sample material from this blended substance and begin assaying the genetic material as per usual.

    This would likely cause a net decrease in the number of kits sold, due to the lowered practicality of sample collection. However, it would make the data infinitely more valuable to partners (assuming the current method is flawed). A good compromise may be just one stool sample instead of 3. The test of accuracy would then be whether the data remains constant from day to day (assuming constant diet and supplementation).

    A fairly simple solution to implement, assuming US mailing laws would not prohibit that size of sample being sent in the post.

    But we can wait for the AmGut results before being too sure of anything 🙂
    – Karl (Elixa)

    • Richard Nikoley on November 10, 2015 at 06:26


      Tim told me in an email last night that AmGut is taking up to a YEAR now, so we won’t be holding our breath.

    • Karl S on November 11, 2015 at 04:22

      For everyone else interested:
      Further to that convo, it transpires (from someone who has a lot more insight on this than I do) that the problem may be in the sequencing methods used. Which, if fundamental, is a shame.

      – Karl (Elixa)

  3. Karl S on November 14, 2015 at 10:12

    Here’s a chart of how Richard’s Lactobacillus genus varied according to uBiome’s JSON data.

    My impressions of this are that
    1. there is no apparent trend.
    2. there is no apparent consistency in results you would expect to remain consistent.

    The first two values vary significantly, and this is prior to any external intervention (e.g. probiotics). The middle two vary significantly also.

    Taking so many tests was fortunate because it has prevented us from drawing an incorrect conclusion. To illustrate how that could have happened, consider the following:

    If we had taken only one test before Elixa and just one on day 12 (i.e. Sept 24th and Oct 7th). It would display a huge increase in the Lactobacillus genus.

    Conversely, if we had taken only the one on Sept 12th and the one on Day 6 (Sept 29th), it would display a decrease in Lactobacilli.

    From this (limited) example, the two logical conclusions that could be made are:
    1. The microbiome varies to this degree from day to day, with insignificant relation to external factors.
    2. The data is not representative of the microbiome.

    I think one of those unlikely – and some will say extremely unlikely – but it’s worth mentioning both to be rigourous.

    Hi Chloe 🙂
    Perhaps their interpretation of the data may have changed, as opposed to the raw data itself.
    The fastq is the raw data which can be analysed on Linux systems (etc) and the JSON is the interpretation of it by uBiome.

    As Tim Steele explains:
    ‘Additionally, uBiome uses a proprietary database of sequences. MG-Rast uses 6 different databases, including Silva, which is regarded as the most complete database in existence.’

    So perhaps uBiome re-evaluated their database and this led to a change in their interpretation (JSON) of the raw data (fastq).

    So it depends if it was the JSON that changed, or the fastq?

    It’s likely they just missed the email if it’s been that long (easy mistake to make!) 🙂

    – Karl (Elixa)

  4. Chloe Archard on November 14, 2015 at 02:11

    I submitted a sample to uBiome in August and received the results back. Second sample submitted in October. When I logged in to compare the data between the two samples the August data had completely changed to what was originally there (I know because I downloaded all the raw data first time around). uBiome haven’t notified me about those changes and I am still waiting to hear back from them 8 days after I emailed them (and received an automated reply saying they aim to get back within 24 hours!)

    If one of the useful things to use uBiome for is to track our own data over time, then we at least expect that data to stay constant in uBiome’s dashboard and not have changed by the next time we log in!!

    No idea what’s going on there. Will update here if I get any clarification from uBiome. But at the moment I am doubting the usefulness of the test…

    • * on February 10, 2017 at 01:42

      I wasn’t so lucky enough to get any results from Ubiome. At least not since I submitted them 3.5 months ago.

  5. Chloe Archard on November 14, 2015 at 02:16

    The only other thing to say was that between the three sets of results I now have (from only two submitted samples), there is good correlation between the genus of bacteria present and those absent, so that at least says something about the test I guess. Would be more worried if I had wildly different bacteria between the different samples. I was missing bifido in all three sets of results 🙁

    • Richard Nikoley on November 14, 2015 at 08:31

      I suspect the lack of bifido is some sort of testing error.

      Tim Steele has ran all my sets through MG-RAST and none have bifido. Zero, zilch. Plenty of lacto.

      This would be highly unlikely in any test subject, but there was zero even after 12 days of Elixa, with each dose having 55 bil CFU of four different bifido strains.

      If counts were very low I’d have reason to suspect some gut problem, but zero smells of testing error to me. Karl concurs.

  6. Chloe Archard on November 17, 2015 at 23:21

    Response from uBiome:

    “We are constantly upgrading our database and bioinformatics algorithms with the latest scientific research data as the field develops more and more – in fact, we recently pushed a huge update. This could definitely shift percentages for you!”

    I have asked them to clarify if it’s just the raw data in JSON format that has changed and whether the raw data in fastq format is still the same as I think they are saying it is their interpretation of the data that has changed.

Leave a Comment

Follow by Email8k