Variable information across SNPs in GWAS data can cause false rejections of colocalisation which can be resolved by proportional colocalisation tests

Fine-mapping is now a standard post-GWAS analysis, but it has been shown to be potentially inaccurate for large meta-analysis GWAS. We show how this can be caused by variable amounts of statistical information between variants, e.g. due to variable sample sizes, which arise widely in meta-analyses, or due to variable imputation accuracy or call rate. This effect becomes more pronounced as sample sizes become larger, such that small differences in information available at different variants can lead to large errors in credible set predictions. Here, we show this inaccuracy propagates to the colocalisation method coloc, yielding confident but incorrect inference. We propose an adaptation of the proportional colocalisation test, which avoids any fine-mapping inference as an alternative to fine-mapping based colocalisation. This method was originally superseded by coloc because the null hypothesis is colocalization, and because it can have low power in genetic regions with many variants. Our adaptation resolves this second issue, and simulations demonstrate that our approach accurately resolves false coloc inferences. However, the first issue remains, such that we suggest proportional colocalisation is used as an accompaniment to fine-mapping-based coloc. Our findings also emphasize the need for new methodologies in fine-mapping, particularly as GWAS sample sizes grow, to ensure accurate fine-mapping inference.