I would like to perform BOLT-LMM to assess the association of SNP with a quantitative trait. BOLT-LMM requires PLINK format files (bed, bim, fam) and imputed files (bgen, sample). Should I use direct genotype data typed from SNP arrays for the plink format files? Or should I use the imputed genotype data converted to plink format?

Also, do PLINK format files require quality control (e.g., removal of SNPs with MAF<0.01)? Please let us know.