As mentioned in the sample genotyping efficiency section above, marker genotyping efficiency (the proportion of samples with a genotype call for each marker) is a good indicator of marker quality. SNP assays that failed on a large number of samples are poor assays, and are likely to result in spurious data. A recommended threshold for removing SNPs with low call rate is approximately 98-99%, although as mentioned in the sample genotyping efficiency section, this threshold may vary from study to study. Marker genotyping efficiency can be reviewed using the --missing option in PLINK. We recommend removing poor quality SNPs before running the sample genotyping efficiency check discussed above, so that fewer samples will be dropped from the analysis simply because they were genotyped with SNP assays that had poor performance (see Figure 6). Markers can be removed based on call rate by using the --geno option, followed by a threshold for a lower limit of missingness (e.g., --geno 0.02 would remove SNPs with more than 2% missing, i.e. less than a 98% call rate).
