A Network Medicine Approach to Investigating ME/CFS Pathogenesis in Severely Ill Patients: A Pilot Study, 2024, Hung, Davis, Xiao

[forestglip]

Maybe the statistics aren't that complicated, since we're just looking at a single variant and not collapsing multiple variants in a gene.

For this one specific variant, a Fisher's exact test comparing allele count in this study (7/40) to total allele count in gnomAD (94/1613706) gives an odds ratio of 3641.3 and a p-value of 5.66e-23.

As far as the number of variants they looked at, from methods:

High-confidence variants were derived by applying a series of filters as follows:

• Variants had to be supported by reads from both strands.
• Conflicting variants (i.e., multiple types of variants were called at the same position) were excluded.
• Variants located within repetitive regions were excluded.
• Variants in the proximity of miscalls were removed to eliminate sequencing errors in specific regions.
• To reduce miscalls resulting from in-vitro contaminations or other systematic errors, a reasonable variant must have at least one homozygous sample in which both alleles are consistent with the reference.

In total, 2,798,019 high-confidence variants passed these five filters and were subsequently used in the following variant analysis.

I think we can do multiple test correction based on 2,298,019 tests. But I don't think the specific number matters much here because of how significant the finding is. Wikipedia says there are around 324 million known human variants. Even if we assume they tested every human variant and we did Bonferroni correction on the above p-value for 324 million tests, that would still leave the corrected p-value extremely significant: Bonferroni p=1.83e-14

(I'm not sure if any of the 7 individuals with the variant are related. That'd make the finding less significant, though it could still be impressive depending on how many individuals are related.)

I think we need to allow for the possibility of genotyping error at this variant, because of how big of a deal this would be if it was real. It'd be a cause identified in a significant proportion of people with severe ME/CFS.

I would think it'd be relatively easy and a major priority to validate this specific variant in other cohorts. Considering how rare the variant is in the general population, I think if even 1 out of 50 other people with severe ME/CFS had this variant (and we were sure the sequencing result in the follow-up wasn't an error), that'd be good evidence of this being real.