Exploration of weighting schemes based on allele frequency and annotation for weighted burden association analysis of complex phenotypes

Weighted burden analysis can incorporate variants with different frequencies and annotations into a combined test for association between a gene and a phenotype. However there has not been a systematic exploration of which weighting schemes provide maximum power to detect association. Here we assess different weighting schemes using a number of genes for which exome-wide evidence of association with common phenotypes was obtained in 200,000 exome-sequenced UK Biobank participants. We find that there are marked differences in optimal weighting schemes between genes, both with respect to allele frequency and to annotation, implying that there is no one-size-fits-all scheme which is generally optimal. It seems helpful to weight rare variants more highly than common ones, to give loss of function variants higher weights than protein-altering variants and to assign higher weights to protein-altering variants predicted to have more severe effects. However with the data currently available it does not seem possible to make more specific recommendations. This research has been conducted using the UK Biobank Resource.

Automated summary

Weighted burden analysis combines variants with different frequencies and annotations for association testing between genes and phenotypes. The optimal weighting schemes differ between genes, indicating there is no universally optimal scheme. Rare variants should be weighted more heavily, loss of function variants should have higher weights, and protein-altering variants with more severe effects should be given higher weights.