Thread by @jakphd

Suyash Shringarpure (@suyashss) from @23andMeResearch presented a fantastic #raredisease study: "Novel genetic associations for rare diseases with GWAS and trans-ethnic analysis of self-reported medical data"

#ASHG21

medrxiv.org/content/10.110…

.@suyashss: It's well known that self-reported data works very well for common diseases, but what about rare diseases? The assumption is that it wouldn't work.

The other common assumption is that rare disease requires sequencing to find rare causal variants.
#ASHG21

.@suyashss: They launched a survey to evaluate these two commonly held assumptions about genetics in #raredisease.

#ASHG21

.@suyashss: To evaluate their self-reported data, they tested whether GWAS could re-discover known genetic associations with rare diseases (e.g., Huntington's disease, cystic fibrosis) and did, even with sample sizes as low as 41 cases (in Huntington's disease)!
#ASHG21

.@suyashss: As an additional validation test, they compared effect sizes from their GWAS to published studies of these diseases (which themselves mostly have different cohort compositions) and effect sizes were very consistent.
#ASHG21

.@suyashss: With these validation analyses completed, they looked for novel associations and found 5.

1 novel association for vestibular schwannoma near CDKN2A/CDKN2B locus.

#ASHG21

.@suyashss: 3 novel associations for spontaneous pneumothorax. Through proximity and eQTLs point, these loci point to 3 genes: KY, POSTN, CFDP1 (all of which are linked to lung function)

#ASHG21

.@suyashss: Duane retraction syndrome (eye-movement disorder) found a locus near the OLIG1 gene. Conditional analysis found a second locus near OLIG2.

#ASHG21

.@suyashss: They attempted to replicate all 5 novel loci. 3 loci (1 in each trait) replicated, but all 5 have the same direction of effect in the replication data. Smaller sample size in replication data may have hurt the replication chances.

#ASHG21

.@suyashss: These prior analyses were all done in individuals of European ancestries and excluded related individuals.

Using @weizhouw's SAIGE method, they found better p-values in a trans-ancestry analysis for 3 of their loci (these were the same 3 loci that replicated)
#ASHG21