Interesting, thanks Veera. With respect to the specific data on TYK2 and infection, do we not just think this is a power issue. Infection is rare, and I would be worried about the power of these approaches in the linked paper in PLOS Med. Although I buy the principle!
Thank you for your comment. To clarify, you meant that the reason we don't see risk associations with infections is because we don't have enough statistical power to find them?
Yes, sorry, I should have been more clear. We have a bit more power now (in UKB, in 2023), but certainly in the mentioned study (https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0122271), the size of the cohort would preclude identifying anything but massive effects. Look at the effect of IL-6 inhibition in COVID-19 using the sentinel rs2228145 variant in COVID-19 HGI. This is a large effect - I mean, tocilizumab reduces mortality by ~8% (absolute!!) but the p val is (if i recall correctly) around 1 x 10-4 even in the latest COVID-19 severe GWAS (~20,000 cases, > 1 million controls). And COVID-19 is a phenotypically simple infection! For more complex infections I think sample sizes are an issue.
Interesting, thanks Veera. With respect to the specific data on TYK2 and infection, do we not just think this is a power issue. Infection is rare, and I would be worried about the power of these approaches in the linked paper in PLOS Med. Although I buy the principle!
Thank you for your comment. To clarify, you meant that the reason we don't see risk associations with infections is because we don't have enough statistical power to find them?
Yes, sorry, I should have been more clear. We have a bit more power now (in UKB, in 2023), but certainly in the mentioned study (https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0122271), the size of the cohort would preclude identifying anything but massive effects. Look at the effect of IL-6 inhibition in COVID-19 using the sentinel rs2228145 variant in COVID-19 HGI. This is a large effect - I mean, tocilizumab reduces mortality by ~8% (absolute!!) but the p val is (if i recall correctly) around 1 x 10-4 even in the latest COVID-19 severe GWAS (~20,000 cases, > 1 million controls). And COVID-19 is a phenotypically simple infection! For more complex infections I think sample sizes are an issue.
Thanks, Fergus! I see your point.