To integrate the fine-mapping outputs provided by FinnGen, we had to take a different approach. Each "locus" in FinnGen is a genomic region that may contain multiple independent causal variants. SuSIE is run on each entire locus, and where there are likely to be multiple causal variants, the SuSIE output provides the probability for each variant to be in a credible set. (Note that for simplicity we use only the SuSIE results from FinnGen, and not FINEMAP results.) We use these variant probabilities from SuSIE credible sets in downstream applications, such as locus-to-gene scoring. However, for display of associated loci, we chose to represent each locus by the single variant with the lowest p-value. This is necessary because we do not have conditionally independent summary statistics in FinnGen (as we computed manually for other GWAS studies), and our co-localisation procedure would depend upon these for secondary signals at a locus.