The collaborators have provided me with whole-genome sequencing (30x) called SNPs of ~800 human subjects, processed with a recent version of illumina CASAVA. The SNP markers in VCF posses these quality scores: QUAL, GQ and read depth. I would like to bring these variants to a downstream association study with quantitative traits, using an additive allelic model. Since selecting different thresholds for filtering markers based on QUAL/GQ and Depth will lead to a variable loss of genotype calls especially for low coverage heterozygote markers, and in the other hand, miscalling the heterozygotes (REF/ALT) as ALT/ALT will ONLY impact the beta estimates (not the p-value), what would be a justified filtering strategy?
EDIT: The only paper I have found so far which has exploited NGS in a quantitative-traits association pipeline is this; does there exist anything more? link