Genetics

This post covers the variant-to-gene half of Upcott and colleagues' 2026 medRxiv preprint, which paired the MS susceptibility GWAS with blood and brain eQTL data to name the genes that risk alleles act through. It explains what summary data-based Mendelian randomization tests and the thresholds used, the specific confound the HEIDI test exists to catch when two causal variants sit in linkage disequilibrium, why Bayesian colocalization against single-cell atlases does double duty by also assigning each gene to a cell type, the 47 of 387 variants that had no usable proxy in the scoring cohort, the 240 eGenes of which 162 are new to MS, and the three external post-mortem and CSF datasets used to check whether the list means anything in real MS tissue.