Uncurated Microbiome Pipelines in Microbiome Testing Companies

The typical pipeline processing finds the closest match of 16s or shotgun to a reference library. Conceptually this is fine, but when the closest match is not a bacteria found in humans (or very rarely), then the match may have zero value. It is speculative information for information sake.

Some examples from the taxon reported in retail microbiome reports:

  • Sharpea azabuensis:  isolated from the faeces of thoroughbred horses in 2008
  • Olsenella timonensis:isolated in 2015
  • Phoenicibacter congonensis: isolated in 2019 from a Pygmy
  • Gillisia: From marine environment
  • Macrococcus: Found in food animals
  • Clostridium chauvoei:  causative agent of blackleg, a wide spread serious infection of cattle and sheep with high mortality

At the very least, the matching should be done to those reported in humans.

This creates a challenge for the clinician — there is no literature on these bacteria.

But to the capable statistician…

They can often be very useful for determining odds ratios for a specific condition or general good health. A suitably large dataset is needed (thousand of samples). This leaves the clinician between the rock (no literature or studies) and a hard place (“magical” statistical odds ratios).

Reflections on Using Odds Ratios

Recently, I released Odds Ratio–based suggestions and reports, such as the example shown here.

The underlying concepts are straightforward:

  • If you are in an unhealthy range, the goal is to move out of that range.
  • If you are not in an unhealthy range but fall outside a defined healthy range, the goal is to move into the healthy range.

For healthy ranges, interpretation is simple—you are either too high or too low, so the direction of adjustment is clear.

  • If the range is 10%ile to 30 %ile and you are at 29%ile, should you increase or decrease?
  • If the range is 70%ile to 90 %ile and you are at 85%ile, should you increase or decrease?

Unhealthy ranges, however, are more nuanced.

For example:

  • If the unhealthy range is the 10th to 30th percentile and your value is at the 29th percentile, should you increase or decrease?
  • If the unhealthy range is the 70th to 90th percentile and your value is at the 85th percentile, should you increase or decrease?

The core issue is understanding why the range is considered unhealthy—specifically, whether the problem arises from having too little of something or too much.

  • Lactobacillus delbrueckii
    • Unhealthy: 0 – 45%ile
    • Healthy: 48%- 72%ile
  • Lactobacillus gasseri (A Surprise!): 0 -100%ile is Unhealthy
  • Lactobacillus johnsonii: 0 – 72%ile Healthy
  • Lactobacillus taiwanensis: 23-56%ile Unhealthy
  • Limosilactobacillus reuteri: 0-52%ile Unhealthy

The above numbers are slightly suspect because many are P > 0.002, so may not be significant.

ConfidenceBacteria this or less significantBacteria More Significant
P < 0.0012081 Ranges694 Ranges
P < 0.00012383 Ranges392 Ranges
P < 0.000012495280

A List of Families with high significance

It is interesting to note that neither Lactobacillaceae (Lactobacillus) nor Bifidobacteriaceae (Bifidobacterium) were found to be significant at this level. Nor were they at the genus level, only at the specific species level (see below)

tax_nameRangeNature
Anaplasmataceae0 to 75Unhealthy
Bartonellaceae0 to 79Unhealthy
Chlorobiaceae0 to 58Unhealthy
Chrysiogenaceae0 to 83Unhealthy
Clostridiales Family XVI. Incertae Sedis0 to 100Unhealthy
Comamonadaceae0 to 28Unhealthy
Comamonadaceae34 to 52Unhealthy
Cyanobacteriaceae21 to 99Unhealthy
Deinococcaceae0 to 46Unhealthy
Desulfonatronaceae0 to 50Unhealthy
Enterococcaceae0 to 66Healthy
Enterococcaceae68 to 71Unhealthy
Euzebyaceae13 to 100Unhealthy
Hyphomicrobiaceae0 to 87Unhealthy
Kiloniellaceae20 to 99Unhealthy
Legionellaceae0 to 100Unhealthy
Listeriaceae74 to 88Unhealthy
Litorivicinaceae0 to 18Unhealthy
Lysobacteraceae0 to 27Unhealthy
Methylophilaceae0 to 96Unhealthy
Nostocaceae14 to 88Unhealthy
Oxalobacteraceae0 to 30Unhealthy
Pseudanabaenaceae0 to 100Unhealthy
Shewanellaceae0 to 73Unhealthy
Sporolactobacillaceae0 to 100Unhealthy
Streptosporangiaceae53 to 88Unhealthy
Symbiobacteriaceae0 to 81Unhealthy
Synechococcaceae0 to 9Unhealthy
Synechococcaceae48 to 85Unhealthy
Thermoanaerobacterales Family III. Incertae Sedis3 to 50Unhealthy
Thiotrichaceae7 to 100Unhealthy
Weeksellaceae0 to 100Unhealthy

Significant Bifidobacterium Species

tax_nameRangeNature
Bifidobacterium adolescentis0 to 49Unhealthy
Bifidobacterium angulatum0 to 73Healthy
Bifidobacterium breve0 to 76Unhealthy
Bifidobacterium catenulatum0 to 100Healthy
Bifidobacterium dentium0 to 50Unhealthy
Bifidobacterium scardovii0 to 79Unhealthy

Significant Lactobacillus Species

tax_nameRangeNature
Lacticaseibacillus brantae33 to 70Unhealthy
Lactobacillus acidophilus0 to 76Unhealthy
Lactobacillus amylovorus0 to 100Unhealthy
Ligilactobacillus murinus0 to 100Unhealthy

Summary

I am hoping to expand the size of my “Healthy Samples” in the next weeks. This should improve the ranges and significance.

Questions about using Odds Ratio

An early user of the Odd Ratio approach (see New Suggestions Approach and Ranges for Healthy and Unhealthy Bacteria) raised concerns about some of the bacteria identified.

For example, Odoribacter denticanis was only 0.002%, Clostridium akagii was 0.002%, and Symbiobacterium was 0.004%. The natural question is whether bacteria present at such tiny levels could have any meaningful impact. Interestingly, these values are not extreme outliers; they appear to be fairly common at these levels.

There are several ways to approach microbiome adjustment. One is to focus on bacteria that dominate the microbiome. Another is to target bacteria with extreme values. A third is to focus on bacteria whose mechanisms of impact are known, such as those that produce metabolites linked to leaky gut.

My own approach is based on strong statistical associations. Association does not prove causation, and in microbiome research, the causal details are often not well established. My working assumption is that bacteria strongly associated with a condition are likely influencing it, perhaps through metabolites they produce or consume. If so, reducing those bacteria should reduce the metabolic effect.

The low-abundance dilemma

The microbiome can be thought of as a population, much like a country’s human population. If there were 989 billionaires[Forbes] in the United States, that would still be only about 0.0003% of the population. Yet few people would conclude from that alone that billionaires have little influence on the country. In practice, a very small number of highly influential actors can still shape outcomes in major ways.

The same logic applies to microbiome analysis. Low abundance does not necessarily mean low impact.

The odds ratios used here are not based on an ideal dataset, but on the best data currently available. The choice is not between perfect evidence and flawed evidence; it is between using the best evidence now or waiting indefinitely for perfect data. In that sense, this is a best-effort approach grounded in the data we have rather than silence in the face of incomplete evidence.

Reader Response

I think the question is whether such low values represent an actual organism or noise. I remember in the days of Ubiome, a reading of 0.001% meant only a single organism was found. Ubiome actually discarded it if there was only one found. They only reported if there were two or more. Thryve otoh, reported everything, which is one of the reasons they found more than Ubiome.

ofc, some of these results are more than one organism, but the question still remains as to whether this is a real organism or noise. It just seems that there are a lot of variables in this analysis with big error margins, and you are compounding them by bundling them all together. The error margin in the final result is likely huge.

Resolution

The way to handle this issue was requiring the raw count to be at least 5. This should reduce the noise level to acceptable levels. The dilemma remains on identification differences between tests (See this post for details). With aggregation across different tests, this issue should be reduced.

New Suggestions Approach

In the decades that I have been working with the microbiome, the scientist in me have become very troubled. Some of the key concerns have been:

  • Massive inconsistency between tests results in terms of percentage of different bacteria found [more information]
  • Medical practitioner picking certain key bacteria to focus on based on rote or hearsay.
    • No studies showing any bacteria are more important than other possible bacteria.
  • Suggestions often do not consider counter-indication / adverse effects on other bacteria
  • Healthy ranges are determined using normal distributions (Normal Range) which is grossly invalid given the typical bacteria distributions [more information]

Microbiome Prescription current suggestion algorithms appears to have over a 75% chance of improving microbiome tests results (with typical subjective improvement reported) [more information]. For those not responding well to those suggestions, I have been researching an alternative, more rigorous, approach based on computationally intense computation. This method is not practical to run on a website, instead it is computed off-line and then emailed to the person.

The new suggestions are based on the following:

  • Using Percentile ranking for better comparison
  • Using Odds Ratios (a rigorous statistical process using P < 0.001) to select bacteria
  • Every suggestion is checked against every selected bacteria to insure no adverse effect

For more details see “A Proposed Model for Clinical Use of the Human Microbiome

The new suggestions set consists of three reports:

  • Focus on the top 20 bacteria associated with being unhealthy
  • Focus on the top 20 bacteria associated with being healthy
  • Focus on the top 40 bacteria associated with both unhealthy and healthy

Selection is based on the statistical significance. Why three? As more and more bacteria are added to the target bacteria, the fewer modifiers are left that does not have adverse effect on some of the bacteria.

A knowledgeable reader raised some very valid questions here: Questions about using Odds Ratio

Example Reports:

Unhealthy Bacteria

The goal is to shift the bacteria outside of the unhealthy range. Often it is to eliminate it, but in other cases it may be just to push it up and outside the range.

  1. Nostocaceae [family] [1162] 47.5 %ile Unhealthy Range [14 – 88], Plan:Decrease
    Signif: 36.74, Odds:-4.34
    Decreases This Bacteria
    • No Substances without adverse effect on other bacteria
  2. Klebsiella [genus] [570] 46.1%ile Healthy Range [1 – 56], Plan:Increase
    Signif: 26.39, Odds:-3.40
    Increases This Bacteria
    • 4 :non-starch polysaccharides @Sugar and similar
    • 2 :Catechol {Catecholamines} @Flavonoids, Polyphenols etc
    • 2 :Escherichia coli:DSM 16441-16448 {symbioflor-2} @Probiotics
    • 2 :Ceratonia siliqua {carob} @Food (excluding seasonings)
    • 2 :115 different soil based taxa {General Biotics Equilibrium} @Probiotics
    • 2 :Laurencia tristicha {Marine red algae} @Food (excluding seasonings)
    • 2 :L-3-hydroxytrimethylaminobutanoate {carnitine} @Amino Acid and similar

Healthy Bacteria

The goal is to shift into the optimal range

  1. Enterococcaceae [family] [81852] 87.2%ile Healthy Range [1 – 66], Plan:Decrease to Healthy Range
    Signif: 20.57, Odds:0.97
    Decreases This Bacteria
    • 7 :Shen Ling Bai Zhu San {参苓白术散} @Herb or Spice
    • 6 :Sambucus nigra L. ssp. canadensis {Elderberry} @Food (excluding seasonings)
    • 5 :Chamaemelum nobile {Camomile} @Herb or Spice
    • 5 :Olea europaea {Olive leaf} @Herb or Spice
    • 5 :Ocimum tenuiflorum {Tulsi} @Herb or Spice
    • 5 :Micromeria fruticosa {White-leaved Savory} @Herb or Spice
    • 5 :Eucalyptus {Gum Tree} @Food (excluding seasonings)
    • 5 :Arctostaphylos uva-ursi {Bearberry} @Food (excluding seasonings)
    • 4 :helichrysum italicum {Immortelle} @Herb or Spice
    • 4 :Dysphania ambrosioides {Epazote} @Food (excluding seasonings)
    • 4 :Sinapis alba {yellow mustard} @Food (excluding seasonings)
    • 4 :Cathelicidin antimicrobial peptide {LL37} @Amino Acid and similar
    • 3 :Bixa orellana {annatto } @Herb or Spice
    • 3 :Rhus coriaria {Sumac} @Herb or Spice
    • 2 :Withania somnifera {Ashwagandha} @Herb or Spice
  2. Streptococcus alactolyticus [species] [29389] 99.0%ile Healthy Range [1 – 75], Plan:Decrease to Healthy Range
    Signif: 17.33, Odds:1.55
    Decreases This Bacteria
    • 5 :Olea europaea {Olive leaf} @Herb or Spice
    • 5 :Micromeria fruticosa {White-leaved Savory} @Herb or Spice
    • 5 :Eucalyptus {Gum Tree} @Food (excluding seasonings)
    • 5 :Sodium Bicarbonate {Baking Soda} @Common and OTC Supplements
    • 5 :Prunus mume {Umeboshi} @Food (excluding seasonings)
    • 4 :helichrysum italicum {Immortelle} @Herb or Spice
    • 4 :chlorhexidine @Common and OTC Supplements
    • 3 :Bixa orellana {annatto } @Herb or Spice
    • 3 :Rhus coriaria {Sumac} @Herb or Spice
    • 2 :Withania somnifera {Ashwagandha} @Herb or Spice
    • 2 :Bifidobacterium longum subsp. longum BB536 {BB536} @Probiotics
    • 2 :Xylaria hypoxylon {candlesnuff fungus} @Food (excluding seasonings)

Both Healthy and Unhealthy

Because the number of bacteria increases, the number of modifiers decrease,

  1. Nostocaceae [family] [1162] 47.5 %ile Unhealthy Range [14 – 88], Plan:Decrease
    Signif: 36.74, Odds:-4.34
    Decreases This Bacteria
    • No Substances without adverse effect on other bacteria
  2. Klebsiella [genus] [570] 46.1%ile Healthy Range [1 – 56], Plan:Increase
    Signif: 26.39, Odds:-3.40
    Increases This Bacteria
    • 4 :Catechol {Catecholamines} @Flavonoids, Polyphenols etc
    • 3 :Escherichia coli:DSM 16441-16448 {symbioflor-2} @Probiotics
    • 3 :L-3-hydroxytrimethylaminobutanoate {carnitine} @Amino Acid and similar
  3. Acidaminococcus [genus] [904] 55.2%ile Healthy Range [21 – 65], Plan:Increase
    Signif: 23.71, Odds:-3.42
    Increases This Bacteria
    • No Substances without adverse effect on other bacteria
  4. Thiotrichaceae [family] [135617] 19.6%ile Unhealthy Range [7 – 100], Plan:Decrease
    Signif: 23.27, Odds:-3.42
    Decreases This Bacteria
    • No Substances without adverse effect on other bacteria
  5. Oscillospira guilliermondii [species] [119853] 67.1%ile Healthy Range [2 – 71], Plan:Increase
    Signif: 22.46, Odds:-3.37
    Increases This Bacteria
    • No Substances without adverse effect on other bacteria
  6. Fusobacterium [genus] [848] 43.3%ile Healthy Range [1 – 70], Plan:Increase
    Signif: 20.81, Odds:-1.67
    Increases This Bacteria
    • 6 :Citricidal {Grapefruit seed extract} @Food (excluding seasonings)
    • 2 :pediococcus acidilactici {P acidilactici} @Probiotics

The Avoid List

This lists the items with the number of bacteria it may adversely impact.

Counter Indicated Modifiers

  • 18 :Slow digestible carbohydrates. {Low Glycemic} @Diet Style
  • 16 :Bovine Milk Products {Dairy} @Food (excluding seasonings)
  • 15 :bacillus @Probiotics
  • 15 :whole-grain diet @Diet Style
  • 15 :bacillus,lactobacillus,streptococcus,saccharomyces probiotic @Probiotics
  • 14 :Hordeum vulgare {Barley} @Food (excluding seasonings)
  • 14 :dietary fiber @Diet Style
  • 14 :High-fibre diet {Whole food diet} @Diet Style
  • 14 :Lacticaseibacillus paracasei {L.paracasei} @Probiotics
  • 14 :Limosilactobacillus reuteri {L. Reuteri} @Probiotics
  • 14 :wheat @Food (excluding seasonings)
  • 14 :yogurt @Food (excluding seasonings)
  • 14 :Fiber, total dietary @Diet Style
  • 13 :bacillus subtilis {B.Subtilis } @Probiotics

How do you get these Reports?

When new suitable samples are uploaded, these reports are automatically generated and email within a day.

If you want an older sample processed, just click this link on the site

Note: If you do not receive it in 36 hours, check your spam and trash folders.

After some user feedback, a single report is sent. This report looks at bacteria in the unhealthy range and those that are outside of the healthy range. Most bacteria has one OR the other.