People who have uploaded their (or their children) samples have annotated some with Autism. On my main blog, I have been doing a series of deep statistical dives on my main microbiome blog site, and checked it the data available reaches the threshold for inclusion as defined in A new specialized selection of suggestions links (A summary table of various studies has been added there).
I am posting here to since it is a blog focused on Autism.
- Bacteria Detected with z-score > 2.6: found 196 items, highest value was 8.2
- Enzymes Detected with z-score > 2.6: found 203 items, highest value was 7.6
- Compound Detected with z-score > 2.6: found ZERO items
For the number of samples, the z-score are unusually high compared to other studies in the series. A high z-score means strong statistical significance.
Interesting Significant Bacteria
All bacteria found significant had too low levels with the exception of the following that was too high:
- Prevotellaceae (family)
- Prevotella copri (species)
- Prevotella veroralis (species)
- Bifidobacterium gallicum (species)
- Peptococcaceae (family)
The most significant ones are listed below. I should point out that these bacteria may not be the cause, rather they may be ‘the canaries in the coal mine’ of the microbiome. These studies’ methodology determines association and not causality.
|Pseudoclostridium thermosuccinogenes (species)||145||70||7|
|Bacteroides rodentium (species)||3371||1481||6.9|
|Anaerovibrio lipolyticus (species)||1328||355||6.7|
|Blautia hydrogenotrophica (species)||252||73||5.7|
|Butyrivibrio proteoclasticus (species)||498||120||5.6|
|Adlercreutzia equolifaciens (species)||232||90||5.5|
|Anaerotruncus colihominis (species)||1833||1140||5.4|
All enzymes found significant had too low levels.
I will leave it to the reader to go to Kyoto Encyclopedia of Genes and Genomes to learn about these enzymes (a steep learning curve).
The top item S-adenosyl-L-methionine (SAMe), had significant literature associated with it, a few samples:
- “S-adenosyl methionine prevents ASD like behaviors triggered by…” 
- “S-Adenosyl-Methionine alleviates sociability aversion and reduces changes in gene expression in a mouse model of social hierarchy” 
- “Blood biomarker levels of methylation capacity in autism spectrum disorder: a systematic review and meta-analysis”  – decrease SAMe in blood
|Enzyme||Reference Mean||Study Mean||Z-Score|
|S-adenosyl-L-methionine:23S rRNA (guanine2535-N1)-methyltransferase (22.214.171.124)||151||70||7.6|
|GDP-alpha-D-rhamnose:NAD(P)+ 4-oxidoreductase (126.96.36.1991)||481||120||6.4|
|ATP:D-galacturonate 1-phosphotransferase (188.8.131.52)||506||120||5.7|
|NAD+:poly(ADP-D-ribosyl)-acceptor ADP-D-ribosyl-transferase (184.108.40.206)||505||120||5.6|
|beta-D-galactosyl-(1->4)-L-rhamnose:phosphate 1-alpha-D-galactosyltransferase (220.127.116.11)||505||120||5.6|
|delta/gamma-tocopherol lyase (ring-opening) (18.104.22.168)||703||213||5.6|
|all-trans-heptaprenyl-diphosphate diphosphate-lyase (cyclizing, tetraprenyl-beta-curcumene-forming) (22.214.171.124)||304||157||4.7|
|CDP-diacylglycerol:myo-inositol 3-phosphatidyltransferase (126.96.36.199)||133||88||4.6|
|Hg:NADP+ oxidoreductase (188.8.131.52)||774||195||4.6|
|ferredoxin:dinitrogen oxidoreductase (ATP-hydrolysing, molybdenum-dependent) (184.108.40.206)||2930||1478||4.4|
|myo-inositol-hexakisphosphate 4-phosphohydrolase (220.127.116.11)||2163||1074||4.2|
|phenol,NADPH:oxygen oxidoreductase (2-hydroxylating) (18.104.22.168)||326||51||4.1|
|1,2-beta-D-glucan:phosphate alpha-D-glucosyltransferase (22.214.171.1243)||264||128||4|
|acetyl-CoA:2-oxoglutarate C-acetyltransferase (thioester-hydrolysing, carboxymethyl-forming) (126.96.36.199)||936||341||4|
Cross Validation To Literature
Using microbiome studies from published studies is often difficult because the results often vary greatly due to the lab equipment used and the software used to interpret the data.
Checking against the literature we found:
- Items reported High
- Items Reported Low – none are reported in the literature
The purpose of these studies is to enable apple-to-apple-to-apple analysis and suggestions. The same lab is used for the raw read (Ombre and Biomesight uses the same lab) and all of the data is processed thru the same software (BiomeSight). Your sample (if you use the website) is processed thru the same process.
A final comment on most published studies, typically there are 25 Autism patients and 25 control patients. This actually results in much lower sensitivity in those studies than we have with over 1000 as the control.
The SAMe “shouting out” in the enzymes being in agreement with published studies causes me to have considerable confidence in the analysis. Dr. Artificial Intelligence suggestions on Microbiome Prescription look like they are likely to have positive results.
P.S. If anyone does more drill down into the enzymes above in the literature, feel free to add your findings as comments.