The research detailed in Connally et al. 2021 (https://www.medrxiv.org/content/10.1101/2021.06.08.21258515v2) requires GWAS summary statistics for non-coding variants without the effects of nearby coding variants.

For GWAS on type II diabetes (Mahajan et al. 2018), Crohn disease (Liu et al. 2015, Goyette et al. 2015), ulcerative colitis (Liu et al. 2015, Goyette et al. 2015), and breast cancer (Zhang et al. 2020), the effects of coding variants were removed from summary statistics using conditional and joint analysis (COJO, Yang et al. 2012). For LDL cholesterol, HDL cholesterol, and height, GWAS were run from UKBB data using coding variants as covariates. The coding SNPs to condition on are selected as described in Connally et al. (2021).

These GWAS do not measure all associations, only those of non-coding variants.