This morning I got this email:

My daughter’s light sensitivity is now so bad, she’s screaming in pain at daylight  and won’t let her flatmate put up the blinds! Of course it’s related to her autism. Now we’ve uploaded her new sample, is there anything implicated in her current dysbiosis that might lessen this?
She is tormented by this..

I believe we just have enough data to get some traction. I will first use the new Odds Ratio because it give an objective measurement of the importance of each bacteria. Second, I will use the older methodology to simply get a second opinion of which bacteria (unfortunately, this does not indicate importance of each bacteria).

There are three symptom choices related. The difference in count is a reflection of when the symptom was added (the earliest one had the highest count).

• Neurological-Vision: photophobia (Light Sensitivity) 431 samples
• DePaul University Fatigue Questionnaire : Abnormal sensitivity to light 259 samples
• Other:Light sensitivity (photophobia) 5

The sample above was done using biomesight and we have 148 different bacteria using Odds that are statistically significant for increasing or reducing the odds.

The Odds of her having light sensitivity is quite high: log(Odds)=11.8,

These notes document ongoing work on this issue. The goal is both to address her request and to deepen our understanding of how the MP classic method compares to the newer Odds Ratio approach. The MP classic method has produced good results so far, and Odds Ratios may further improve them. For details on how Odds Ratios are calculated, see this related post: Odds Ratio for the Microbiome 101.

In subsequent posts I will look at two symptoms that are very often seen with light sensisitivy:

• Multiple Chemical Sensitivity
• Mast Cell Activation Syndrome

Comparison of “MP Classic” and Odds Ratio Algorithms

Across all symptoms, using Biomesight data, we see consistent patterns in which bacterial levels are involved. The Odds Ratio analysis focuses on more specific bacterial taxa and is therefore more targeted. For example, instead of simply indicating low Lactobacillus, the Odds Ratio can highlight a particular species such as Lactobacillus reuteri. This higher resolution enables more precise selection of probiotics.

Taxonomy Rank	MP Classic	Odds Ratio
Species	1727	13541
Genus	5130	10040
Family	8463	6158
Order	5860	3269
Class	3663	1437

Overview of all Samples

The list of bacteria that DOUBLES or more the odds when present in larger amounts

Bacteria	Rank	Odds Ratio
Salidesulfovibrio	genus	5.9
Salidesulfovibrio brasiliensis	species	5.9
Ethanoligenens	genus	4.9
Peptoniphilus lacrimalis	species	4.3
Slackia faecicanis	species	4.2
Collinsella tanakaei	species	3.8
Finegoldia magna	species	3.5
Viviparoidea	superfamily	3.5
Architaenioglossa	order	3.5
Rivularia	genus	3.5
Viviparidae	family	3.5
Rivularia atra	species	3.5
Rivularia	genus	3.5
Finegoldia	genus	3.4
Lysobacter	genus	3.4
Desulfovibrio fairfieldensis	species	3.3
Aerococcaceae	family	3.3
Anaerococcus	genus	3.2
Streptococcus anginosus	species	3.1
Luteolibacter	genus	3
Luteolibacter algae	species	3
Anaerotruncus colihominis	species	3
Odoribacter denticanis	species	3
Filifactor	genus	2.8
Lactobacillus gallinarum	species	2.8
Peptoniphilus asaccharolyticus	species	2.8
Selenomonas infelix	species	2.7
Corynebacterium striatum	species	2.7
Adlercreutzia equolifaciens	species	2.6
Streptococcus anginosus group	species group	2.6
Glutamicibacter soli	species	2.6
Anaerotruncus	genus	2.5
Rubritaleaceae	family	2.5
Rubritalea	genus	2.5
Gardnerella	genus	2.4
Oscillatoriales	order	2.3
Amedibacillus dolichus	species	2.3
Amedibacillus	genus	2.3
Glutamicibacter	genus	2.2
Anaerococcus prevotii	species	2.2
Azospirillum palatum	species	2.2
Eggerthella sinensis	species	2.2
Sphingomonas abaci	species	2.2
Alcanivorax	genus	2.1
Alcanivoracaceae	family	2.1
Haploplasma	genus	2.1
Haploplasma cavigenitalium	species	2.1
Isoalcanivorax	genus	2.1
Isoalcanivorax indicus	species	2.1
Oscillatoriaceae	family	2.1
Selenomonadales	order	2.1
Nisaea nitritireducens	species	2.1
Anaerococcus tetradius	species	2.1
Selenomonadaceae	family	2.1
Lactobacillus acidophilus	species	2.1
Anaerococcus lactolyticus	species	2.1

On the other end, the bacteria that reduces the odds when present in higher amounts are:

Propionibacteriales	order	0.1
Dyadobacter	genus	0.3
Herbaspirillum magnetovibrio	species	0.3
Calditrichia	class	0.4
Calditrichales	order	0.4
Calditrichaceae	family	0.4
Caldithrix	genus	0.4
Calditrichota	phylum	0.4
Desulfitobacteriaceae	family	0.4
Bifidobacterium adolescentis	species	0.4
Bifidobacterium longum	species	0.4

In terms of probiotics, we see some quick observations: good and bad.

• Two Lactobacillus probiotics significantly increases the odds — i.e. AVOID, especially yogurts!
• Two Bifidobacterium species (and the genus as a whole) significantly decreases the odds — TAKE A LARGER DOSAGE.

Looking at this specific sample

We found no lactobacillus at all, and Bifidobacterium adolescentis is too low. Bifidobacterium longum was found but the amount was significant for reducing the risk.

Getting best probiotics via modelling

This is done using the Correlation Coefficient between bacteria from the R2 site (using the lab specific numbers). We focused solely on the bacteria that increased the odds significantly, and then compute the probiotics (based on only the species what Biomesight reports) that will shift them in the right direction.

Tax_name	Impact
Pediococcus acidilactici	4.28
Bacillus amyloliquefaciens group	3.89
Limosilactobacillus vaginalis	2.95
Bifidobacterium	2.5
Enterococcus faecalis	1.73
Bifidobacterium pseudocatenulatum	1.6
Leuconostoc mesenteroides	1.6
Heyndrickxia coagulans (bacillus coagulans)	1.53
Bifidobacterium longum	1.49
Clostridium butyricum	1.46
Lacticaseibacillus paracasei	1.35
Lactococcus lactis	1.33
Bifidobacterium breve	1.28
Lactobacillus helveticus	1.27
Enterococcus faecium	1.24
Bacillus subtilis group	1.16
Lactiplantibacillus plantarum	1.08
Bifidobacterium bifidum	0.96
Bifidobacterium adolescentis	0.84

Taking these same bacteria using the odds ratios and our usual suggestions engine, we get the following as the top suggestions.

Modifier	Net	Take	Avoid
Slow digestible carbohydrates. {Low Glycemic}	37	52	16
dietary fiber	29	45	16
Fiber, total dietary	24	38	14
fruit	22	34	12
fruit/legume fibre	20	32	12
(2->1)-beta-D-fructofuranan {Inulin}	20	23	3
High-fibre diet {Whole food diet}	19	32	13
oligosaccharides {oligosaccharides}	19	26	6
whole-grain diet	18	25	7
Lactobacillus plantarum {L. plantarum}	17	29	12
bifidobacterium	15	16	1
wheat	12	14	2

The Avoids. I noticed that Bofutsushosan is an avoid. This is a promoter of Akkermansia — which was on our avoid probiotics list. There appears to be reasonable consistency although we are using two different sources and mechanism to get these suggestions.

Modifier	Net	Take	Avoid
high-fat diets	-8	3	11
Ganoderma sichuanense {Reishi Mushroom}	-5	1	6
Pulvis ledebouriellae compositae {Bofutsushosan}	-4	0	5
2-aminoacetic acid {glycine}	-4	0	4
Bacteriophages LH01,T4D,LL12,LL5 {PreforPro}	-4	0	4
laminaria hyperborea {Cuvie}	-4	0	4
low protein diet	-4	1	6
D-glucose {Glucose}	-4	1	6
Ferrum {Iron Supplements}	-4	1	5
Ulmus rubra {slippery elm}	-4	2	6
Honey {Honey }	-4	2	6

Going Old School Suggestions

This is done the usual way but we temporarily clear all of the symptoms and then just marked this single symptom. We are wanting to focus solely on this one horrible symptom.

Clicking on this one symptom, we then get 10 bacteria associated

And also suggestions. I note some agreements between the methods:

• Avoids: Honey, Ganoderma sichuanense {Reishi Mushroom},laminaria hyperborea {Cuvie}, etc
• Takes: whole-grain diet, oligosaccharides
• Disagreement: Bifidobacterium Longum – this gets interesting because the Odds Ratio indicate that the amount of Bifidobacterium Longum present was sufficient to reduce the odds to below 1.0

Summary

I generally favor a consensus of recommendations as the safest course. In this case, my impression is that using Odds Ratios leads to better identification of the bacteria involved (10 versus 24 for this sample), with the added benefit of indicating the relative importance of each bacterium. With Odds Ratios, the thresholds for being too high or too low are symptom-specific, rather than some magical universal cutoff that applies to all conditions.

Believing that there is one magic reference range for any bacteria is simply naive and ignoring the data.

I need to do some more refining of the code as well as enhancement to handle multiple symptoms concurrently; in time, this will be added to the sight.

Using Odds Ratio is now available on the site. The video below shows how to access it.

Technical Notes

Doing a low level comparison between the “classic forecast method” and the “Odds Ratio method I generated the table below. The Odds Ratio identified bacteria at a much more at a finer level (species) and most people would interpret that as being more targeted and likely better outcomes.

Measure	Classic	Odds Ratio
Bacteria Considered	115	148
Bacteria in common	20	20
Species	8	57
Genus	22	51
Family	33	21
Order	23	10
Class	14	3

This also implies that only Genus and Species should be considered with Odds Ratio. Statistically this is preferred to reduce the amount of double counting.

Revisiting Suggestions using only Genus and Species with Odds Ratio

The R2 Probiotics are similar. Most probiotics are more challenging to obtain — see this page for known sources. The avoids are:

• Lactobacillus johnsonii
• Akkermansia muciniphila
• Bacillus subtilis

Note: Pediococcus acidilactici and L.Plantarum (positive) mixtures is likely the easiest to obtain.

Tax_name	Impact	Possible Source
Pediococcus acidilactici	4.28	Imagilin / NutriLots
Bacillus amyloliquefaciens group	3.1	only in big mixtures 🙁
Limosilactobacillus vaginalis	1.79	n/a
Bifidobacterium pseudocatenulatum	1.6	only in big mixtures 🙁
Leuconostoc mesenteroides	1.6	Bulk Probiotics / Leuconostoc Mesenteroides Probiotic Powder
Clostridium butyricum	1.46	Many sources
Lacticaseibacillus paracasei	1.35	danactive drink and many others
Lactococcus lactis	1.33	Bulk Probiotics / Lactococcus Lactis Probiotic Powder
Bifidobacterium	1.04

The To Take List

Modifier	Net	Take	Avoid
Slow digestible carbohydrates. {Low Glycemic}	34	47	12
dietary fiber	29	40	11
Fiber, total dietary	23	35	11
fruit	20	30	10
oligosaccharides {oligosaccharides}	20	24	4
High-fibre diet {Whole food diet}	19	29	10
fruit/legume fibre	19	28	9
whole-grain diet	18	24	5
(2->1)-beta-D-fructofuranan {Inulin}	17	18	1
bifidobacterium	12	12	0
Lactobacillus plantarum {L. plantarum}	11	22	11
wheat	11	12	1
3,3′,4′,5,7-pentahydroxyflavone {Quercetin}	10	11	1
Bovine Milk Products {Dairy}	9	13	4
Human milk oligosaccharides (prebiotic, Holigos, Stachyose)	9	10	1
polyphenols	8	12	4

The To Avoid List

high-fat diets	-6	1	7
Honey {Honey }	-5	1	6
Pulvis ledebouriellae compositae {Bofutsushosan}	-4	0	5
2-aminoacetic acid {glycine}	-4	0	4
laminaria hyperborea {Cuvie}	-4	0	4
Vaccinium myrtillus {Bilberry}	-4	0	4
D-glucose {Glucose}	-4	1	6
Sodium Chloride {Salt}	-4	1	5
Ferrum {Iron Supplements}	-4	1	5
Ulmus rubra {slippery elm}	-4	2	6
2-Amino-5-(carbamoylamino)pentanoic acid {Citrulline}	-3	0	3
Lactotransferrin {Lactoferrin}	-3	0	3
Sus domesticus {Pork}	-3	0	3
Ganoderma sichuanense {Reishi Mushroom}	-3	1	4
low protein diet	-3	1	3
Theobroma cacao {Cacao}	-3	2	5