A reader asked about gluten sensitivity profile in an email. Here are the results. The short form for probiotics:

• Bifidobacterium breve
• Bifidobacterium longum
• Bifidobacterium adolescentis
• AVOID LACTOBACILLUS

This document presents the results of statistical analysis on symptoms from viable, self-annotated Biomesight microbiome samples. The methodology for data acquisition is outlined in New Standards for Microbiome Analysis?.​

Tables have been refined to display only genus- and species-level taxa, the 20 most prominent entries per group, and associations achieving statistical significance (P < 0.01).​

The following sections provide the processed data, accompanied by guidance on interpretation and application. Counts of significant bacterial taxa are included, reflecting the application of non-standard but rigorously validated statistical approaches to extensive sample and reference populations, where statistical power derives from dataset scale.

Significance	Genus
p < 0.01	162
p < 0.001	146
p < 0.0001	131
p < 0.00001	116

Averages and Medians

I prefer medians over averages. Medians are the values where half of the people have less and half has more. If the data was a bell-curve, then the values will almost be the same… with bacteria that happens rarely. Look at the bacterua below, we see that for some the average is above and the median below. Should one increase or decrease this bacteria?

tax_name	Rank	Symptom Average	Reference Average	Symptom Median	Reference Median
Bacteroides	genus	27.548	26	24.269	26.905
Faecalibacterium prausnitzii	species	12.695	12.196	11.329	12.474
Roseburia	genus	2.324	2.857	1.809	1.382
Lachnospira	genus	3.173	2.711	1.885	2.302
Oscillospira	genus	2.668	2.345	1.947	2.323
Bacteroides uniformis	species	2.839	2.728	1.565	1.903
Parabacteroides	genus	3.138	2.607	1.719	2.022
Clostridium	genus	2.087	1.851	1.363	1.531
Pedobacter	genus	1.315	0.988	0.552	0.706
Coprococcus	genus	1.13	1.442	0.73	0.604
Bacteroides thetaiotaomicron	species	0.943	1.077	0.464	0.59
Bifidobacterium	genus	0.352	0.955	0.129	0.028
Hathewaya histolytica	species	0.442	0.273	0.154	0.251
Hathewaya	genus	0.442	0.273	0.154	0.251
Ruminococcus bromii	species	1.039	0.783	0.167	0.262
Bacteroides cellulosilyticus	species	1.266	0.839	0.076	0.155
Bilophila	genus	0.415	0.348	0.209	0.285
Bilophila wadsworthia	species	0.393	0.34	0.199	0.262
Dorea	genus	0.329	0.488	0.295	0.242
Bacteroides rodentium	species	0.361	0.393	0.186	0.235

If symptom median is higher than reference median, it means there is more of this bacteria. If lower, then less. This ignores how often the bacteria is seen (we average only over reports). IMHO using average value instead of median will often result in a worse situation for the patient

Bacteria Incidence – How often is it reported

The common sense belief is that if a bacteria is reported more often, then the amount should be higher. This is often not true. The microbiome is a complex thing. In this case two specific probiotic species are seen rarely and thus, supplementation could be inferred.

tax_name	Rank	Incidence Odds Ratio	Chi2	Symptoms %	Reference %
Bifidobacterium breve	species	0.57	8.8	23.6	41.3
Bifidobacterium catenulatum	species	0.6	6.7	21.6	35.9

More or Less often based on Symptom Median All Incidence

This is a little more complex to understand. If we compute the mid point for people with the symptom, then if the bacteria was not involved then half of the reference should be above this value and half below this value. If not, it means that the symptom tends to over or under growth.

tax_name	Rank	Symptom Median	Odds Ratio	Chi2	Below	Above
Thiorhodococcus mannitoliphagus	species	0.002	0.2	37.9	132	27
Cystobacter	genus	0.002	0.21	37.4	131	27
Psychrobacter glacialis	species	0.002	0.36	33.8	675	243
Rickettsia marmionii Stenos et al. 2005	species	0.002	0.36	30.3	393	140
Niabella	genus	0.002	0.38	29	585	224
Viridibacillus neidei	species	0.002	0.39	27	470	182
Thiorhodococcus	genus	0.002	0.43	22.9	579	247
Thermodesulfovibrio thiophilus	species	0.002	0.44	21.5	541	236
Oenococcus	genus	0.002	0.45	20.7	614	275
Thermodesulfovibrio	genus	0.002	0.45	20.1	626	284
Helicobacter suncus	species	0.002	0.46	19.6	765	355
Viridibacillus	genus	0.002	0.5	14.8	488	244
Desulfotomaculum defluvii	species	0.003	0.56	11.6	1017	569
Alkalibacterium	genus	0.003	0.57	10.6	899	514
Sporotomaculum syntrophicum	species	0.003	0.58	10.4	1127	652
Pelagicoccus	genus	0.002	0.58	10.1	842	487
Treponema	genus	0.003	0.58	9.7	593	342
Olivibacter soli	species	0.002	0.57	9.5	457	262
Hydrogenophilus	genus	0.003	0.59	9.5	1133	671
Mycoplasma iguanae	species	0.002	0.58	9.1	458	266

More or Less often based on Reference Median All Incidence

This is like the above, but with a different line in the sand. Instead of the median of those with the condition, we use the median of the reference set.

tax_name	Rank	Reference Median	Odds Ratio	Chi2	Below	Above
Bifidobacterium	genus	0.028	2.37	347.9	1154	2736
Tetragenococcus	genus	0.004	0.44	234.4	1638	719
Bifidobacterium adolescentis	species	0.004	2.03	215.7	1074	2176
Hathewaya histolytica	species	0.2505	0.52	202.8	2568	1345
Hathewaya	genus	0.2505	0.52	202.2	2567	1346
Psychrobacter glacialis	species	0.002	0.36	168.1	675	243
Anaerotruncus	genus	0.1785	0.57	155.7	2439	1383
Caloramator uzoniensis	species	0.006	0.51	153.9	1408	712
Bifidobacterium choerinum	species	0.0055	1.87	151.9	917	1718
Mogibacterium	genus	0.022	0.57	145.4	2115	1195
Methylonatrum	genus	0.004	0.54	145	1627	872
Methylonatrum kenyense	species	0.004	0.54	145	1627	872
Anaerotruncus colihominis	species	0.1705	0.58	143	2415	1403
Hymenobacter xinjiangensis	species	0.007	0.53	137.4	1486	795
Niabella	genus	0.002	0.38	135.9	585	224
Streptococcus australis	species	0.0095	0.57	127.5	1773	1010
Leptolyngbya laminosa	species	0.0045	0.44	125.9	698	304
Leptolyngbya	genus	0.0045	0.44	125.8	701	306
Bifidobacterium longum	species	0.0195	1.73	124	1047	1814
Vagococcus	genus	0.003	0.48	119.9	841	403

More or Less often based on Symptom Median High Incidence

Above we see that many of the top bacteria identified are sparse, that is not reported often. We then restrict them to those that occur above 50% or the time.

tax_name	Rank	Symptom Median Freq	Odds Ratio	Chi2	Below	Above
Clostridium taeniosporum	species	0.003	0.6	8.9	1354	818

More or Less often based on Reference Median High Incidence

Above we see that many of the top bacteria identified are sparse, that is not reported often. We then restrict them to those that occur above 50% or the time.

tax_name	Rank	Reference Median Freq	Odds Ratio	Chi2	Below	Above
Bifidobacterium	genus	0.028	2.37	347.9	1154	2736
Tetragenococcus	genus	0.004	0.44	234.4	1638	719
Bifidobacterium adolescentis	species	0.004	2.03	215.7	1074	2176
Hathewaya histolytica	species	0.2505	0.52	202.8	2568	1345
Hathewaya	genus	0.2505	0.52	202.2	2567	1346
Psychrobacter glacialis	species	0.002	0.36	168.1	675	243
Anaerotruncus	genus	0.1785	0.57	155.7	2439	1383
Caloramator uzoniensis	species	0.006	0.51	153.9	1408	712
Bifidobacterium choerinum	species	0.0055	1.87	151.9	917	1718
Mogibacterium	genus	0.022	0.57	145.4	2115	1195
Methylonatrum	genus	0.004	0.54	145	1627	872
Methylonatrum kenyense	species	0.004	0.54	145	1627	872
Anaerotruncus colihominis	species	0.1705	0.58	143	2415	1403
Hymenobacter xinjiangensis	species	0.007	0.53	137.4	1486	795
Niabella	genus	0.002	0.38	135.9	585	224
Streptococcus australis	species	0.0095	0.57	127.5	1773	1010
Leptolyngbya laminosa	species	0.0045	0.44	125.9	698	304
Leptolyngbya	genus	0.0045	0.44	125.8	701	306
Bifidobacterium longum	species	0.0195	1.73	124	1047	1814
Vagococcus	genus	0.003	0.48	119.9	841	403

Summary

A large number of bacterial taxa exhibit shifts with P < 0.01 in association with this condition. The subsequent challenge is determining how to modulate these taxa, since the volume of candidates exceeds what most individuals can practically consider. Moreover, for many of the taxa identified, there is no published evidence in the U.S. National Library of Medicine describing how to alter their abundance.

A deep optimization model, such as the one implemented on the Microbiome Taxa R2 site, can be used to inform probiotic selection. This model provides coverage for each identified taxon and infers which probiotics are most likely to shift their levels. Its output may then be integrated with more conventional recommendations derived from literature indexed in the U.S. National Library of Medicine where such evidence exists, with the two recommendation sets reconciled by giving priority to probiotic-based suggestions.

Development of a dedicated database based on Biomesight samples is in progress. The current model uses data contributed by PrecisionBiome, and datasets generated with differing laboratory processing pipelines cannot be safely combined, as discussed in The taxonomy nightmare before Christmas…. Once the Biomesight-specific database is complete, an option for generating (offline-only) personalized suggestions will be added to the Microbiome Prescription website.

Probiotics Suggestions

The following are based on a simplified algorithm using R2 data for Biomesight. These are tentative numbers subject to future refinements. Bacteria listed are only for probiotics detected with Biomesight tests. Probiotics include some that are available only in some countries and some that are pending approval for retail sale.

• Good Count: Number of bacteria expected to shift in desired direction
• Bad Count: Number of bacteria expected to shift in wrong direction
• Impact: Estimator of impact based on Chi-2, Slope and R2 vectors

Probiotic Species	Impact	Good Count	Bad Count
Bifidobacterium breve	103.8	16	0
Bifidobacterium longum	93.37	17	2
Bifidobacterium adolescentis	68.66	14	2
Enterococcus faecalis	42.16	59	24
Lactobacillus johnsonii	30.97	46	20
Segatella copri	24.16	6	0
Bifidobacterium bifidum	21.18	12	3
Bifidobacterium catenulatum	18.93	10	0
Akkermansia muciniphila	15.46	8	19
Lactobacillus helveticus	10.82	46	44
Bifidobacterium animalis	9.67	7	0
Pediococcus acidilactici	6.74	31	20
Enterococcus faecium	6.47	21	26
Blautia wexlerae	2.6	2	1
Streptococcus thermophilus	2.28	7	1
Escherichia coli	2.23	4	2
Clostridium butyricum	2.07	21	3
Enterococcus durans	0.62	27	16
Lactococcus lactis	0.47	8	2
Bacillus subtilis	0.24	14	27
Bifidobacterium pseudocatenulatum	0.17	18	12
Limosilactobacillus fermentum	0.11	8	8
Veillonella atypica	0.07	14	3
Heyndrickxia coagulans	0.01	9	12
Limosilactobacillus reuteri	-0.01	12	13
Lacticaseibacillus paracasei	-0.03	4	10
Lactobacillus crispatus	-0.04	13	13
Ligilactobacillus salivarius	-0.04	1	6
Leuconostoc mesenteroides	-0.09	8	8
Lacticaseibacillus casei	-0.13	2	9
Lactiplantibacillus plantarum	-0.13	0	6
Lactiplantibacillus pentosus	-0.14	0	6
Lacticaseibacillus rhamnosus	-0.17	1	8
Lactobacillus acidophilus	-0.21	16	22
Limosilactobacillus vaginalis	-0.31	32	29
Odoribacter laneus	-0.45	0	3
Lactobacillus jensenii	-1.03	30	26
Blautia hansenii	-3.61	0	14
Parabacteroides goldsteinii	-38.45	0	16
Parabacteroides distasonis	-48.84	0	11
Bacteroides uniformis	-147.38	2	10
Bacteroides thetaiotaomicron	-157.07	2	8
Faecalibacterium prausnitzii	-278.2	2	7