Our use of a heritability-based approach has advantages but also leads to some limitations. First, our approach will not detect strong but highly localized signals. Second, power increases only modestly with sample size at very large sample sizes (Supplementary Note). Also, because our approach uses stratified LD score regression, it cannot be applied to custom array data, it requires a sequenced reference panel that matches the population studied in the GWAS, and can be affected by model misspecification7. Recent augmentations to the baseline model65 have been shown to help ameliorate model misspecification, but we leave further investigation of this in the context of cell-type-specific analyses to future work.
