A reader wrote asked about apparent inconsistencies in suggestions. That is a very valid request.

What’s the difference between ‘General suggestions’ on the odds ratio suggestions page and ‘Probiotic suggestions’? The results seem almost unrelated. Two examples: in my results, Bifidobacterium breve is in the ‘General suggestions’ listed as having a negative impact, yet in ‘Probiotic suggestions’ it has a highly positive impact, is even mentioned as the second best probiotic (see image down below)! The one there mentioned as the top probiotic, Enterococcus faecalis, is in my ‘General suggestions’ hardly a suggestion, having barely any positive impact (0.87, with many probiotics having more impact).

Clinical Studies Based versus Modelled Probiotics

When I saw an opportunity to model the impact of probiotics instead of relying on published studies, I jumped at the opportunity. The key reasons are below:

• Clinical studies often given contradictory results on the impact of a probiotic on other bacteria.
  • There are many reasons that this would occur naturally:
    • The studies were in the context of an existing condition (thus altered microbiome to start with)
    • The studies used different reference libraries to determine bacteria (See Nightmare post)
    • The studies usually gives a simple increase or decrease.
  • Example: for Pseudomonas, we have 2 studies saying it is increased and 2 studies saying it is decreased by B. subtilis
• Modelled uses:
  • Healthy individuals for modelling (thus no existing conditions!)
  • The same reference library for all samples
  • The model gives a numeric estimate of how much changes is expected (R²)

What is the practical implementation? Looking at the differences below (The full tables are bottom) we see that the model shows impact on a magnitude more of different bacteria.

• In theory, over time, with enough studies on healthy individual with sufficient size of each study, there will likely be convergence of the numbers. Studies are time consuming to do — so these results are likely not likely to be fully available until the next millenium.

Probiotic	Model: Different Bacteria impacted	Studies: Different Bacteria impacted
Bacillus pumilus	1272	18
Aspergillus oryzae	1093	40
Lactobacillus kefiranofaciens	1065	42
Enterococcus faecium	1002	99
Lentilactobacillus kefiri	841	26
Bacillus licheniformis	835	77
Faecalibacterium prausnitzii	671	16
Debaryomyces hansenii	613	8
Blautia wexlerae	599	4
Lentilactobacillus buchneri	539	10
Lactobacillus jensenii	491	11
Bacteroides uniformis	481	5
Lactiplantibacillus argentoratensis	446	9
Akkermansia muciniphila	474	43
Heyndrickxia coagulans	499	72
Lactobacillus delbrueckii subsp. bulgaricus	449	34
Blautia hansenii	342	1
Parabacteroides goldsteinii	338	2
Parabacteroides distasonis	310	1
Lactococcus lactis	330	36
Lactobacillus acidophilus	381	89
Latilactobacillus sakei	296	34
Bifidobacterium animalis subsp. lactis	261	81
Pediococcus	214	46
Pediococcus pentosaceus	211	43
Shouchella clausii	173	15
Lactococcus lactis subsp. lactis bv. diacetylactis	158	8
Lactobacillus helveticus	201	61
Escherichia coli	180	42
Lactobacillus johnsonii	179	43
Levilactobacillus brevis	157	23
Streptococcus thermophilus	178	69
Ligilactobacillus salivarius	171	62
Lacticaseibacillus casei	202	112
Limosilactobacillus reuteri	198	116
Bifidobacterium adolescentis	103	38
Lactiplantibacillus pentosus	91	26

The Personal Decision to be made

Microbiome Prescription generates suggestions using both methods. There is no mechanism to determine which is better. Personally, I prefer the model because there is a lot more data available and the data is quantitative and not a binary of (increase/decrease).

The model assumes this logic:

• If you take a (living) probiotic, then the amount in your microbiome will increase and all of the cascading impacts of this increase will likely match the impact of healthy individuals who naturally have more of that bacteria.

Why Contradictions?

There are massive interactions occurring. If you ignore (or have no data) on some impacts, then you can easily go very off course.

Consider Lactobacillus brevis: Assuming that 10% of your bacteria are out of wack, the recommendation with studies would be based on just 2 bacteria (10% of 23). Recommendations from the model would use around 16 bacteria (10% of 157). In short, more factors would be considered.

Appendix: Full Tables

Probiotics	Different Bacteria Impacted according to studies
Lactiplantibacillus plantarum	198
Saccharomyces boulardii (nom. inval.)	170
Lacticaseibacillus rhamnosus	136
Bacillus subtilis group	128
Limosilactobacillus reuteri	116
Lacticaseibacillus paracasei	116
Lacticaseibacillus casei	112
Bifidobacterium longum	106
Bifidobacterium animalis	101
Enterococcus faecium	99
Limosilactobacillus fermentum	96
Lactobacillus acidophilus	89
Bifidobacterium bifidum	86
Bifidobacterium animalis subsp. lactis	81
Bacillus licheniformis	77
Clostridium butyricum	74
Heyndrickxia coagulans	72
Streptococcus thermophilus	69
Bacillus amyloliquefaciens group	63
Ligilactobacillus salivarius	62
Lactobacillus helveticus	61
Lactobacillus gasseri	52
Lacticaseibacillus rhamnosus	48
Brevibacillus laterosporus	48
Pediococcus	46
Pediococcus pentosaceus	43
Akkermansia muciniphila	43
Lactobacillus johnsonii	43
Lactobacillus kefiranofaciens	42
Escherichia coli	42
Aspergillus oryzae	40
Bifidobacterium adolescentis	38
Lactococcus lactis	36
Latilactobacillus sakei	34
Lactobacillus delbrueckii subsp. bulgaricus	34
Bifidobacterium breve	33
Limosilactobacillus mucosae	31
Lacticaseibacillus casei	30
Streptococcus thermophilus	28
Lactiplantibacillus pentosus	26
Lentilactobacillus kefiri	26
Bifidobacterium longum	26
Bifidobacterium longum subsp. infantis	24
Levilactobacillus brevis	23
Enterococcus faecium	23
Bacillus subtilis group	20
Bacillus pumilus	18
Bifidobacterium pseudocatenulatum	17
Faecalibacterium prausnitzii	16
Shouchella clausii	15
Bacillus amyloliquefaciens	14
Enterococcus durans	12
Lactobacillus jensenii	11
Lentilactobacillus buchneri	10
Bifidobacterium catenulatum	10
Escherichia coli	10
Lactiplantibacillus argentoratensis	9
Lactococcus lactis subsp. lactis bv. diacetylactis	8
Debaryomyces hansenii	8
Bacteroides uniformis	5
Escherichia coli	5
Blautia wexlerae	4
Streptomyces albogriseolus	3
Parabacteroides goldsteinii	2
Parabacteroides distasonis	1
Blautia hansenii	1

Compared to the modelled data

Modelled Probiotic	Different Bacteria Impacted according to model
Bacillus pumilus	1272
Aspergillus oryzae	1093
Lactobacillus kefiranofaciens	1065
Enterococcus faecium	1002
Lentilactobacillus kefiri	841
Bacillus licheniformis	835
Faecalibacterium prausnitzii	671
Debaryomyces hansenii	613
Blautia wexlerae	599
Lentilactobacillus buchneri	539
Heyndrickxia coagulans	499
Lactobacillus jensenii	491
Bacteroides uniformis	481
Akkermansia muciniphila	474
Lactobacillus delbrueckii subsp. bulgaricus	449
Lactiplantibacillus argentoratensis	446
Lactobacillus acidophilus	381
Blautia hansenii	342
Parabacteroides goldsteinii	338
Bacillus	331
Lactococcus lactis	330
Parabacteroides distasonis	310
Latilactobacillus sakei	296
Bifidobacterium animalis subsp. lactis	261
Pediococcus	214
Pediococcus pentosaceus	211
Lacticaseibacillus casei	202
Lactobacillus helveticus	201
Limosilactobacillus reuteri	198
Escherichia coli	180
Lactobacillus johnsonii	179
Streptococcus thermophilus	178
Shouchella clausii	173
Ligilactobacillus salivarius	171
Lactococcus lactis subsp. lactis bv. diacetylactis	158
Levilactobacillus brevis	157
Bifidobacterium animalis	116
Bifidobacterium longum	116
Bifidobacterium adolescentis	103
Bifidobacterium breve	96
Limosilactobacillus mucosae	92
Lactiplantibacillus pentosus	91
Bifidobacterium	89
Limosilactobacillus fermentum	88
Bifidobacterium longum subsp. infantis	74
Enterococcus durans	61
Bifidobacterium catenulatum	52
Lactiplantibacillus plantarum	45
Lactobacillus gasseri	35
Bifidobacterium bifidum	29
Bifidobacterium pseudocatenulatum	28
Lacticaseibacillus rhamnosus	28
Clostridium butyricum	18
Lacticaseibacillus paracasei	7