further analysis. Remaining samples were genotyped for all probesets (stage 2 genotyping), using APT. The Affymetrix SNPolisher v1.5.2 software was used to classify probeset quality and determine the best probeset for each marker. Markers whose best probeset classified as “Other”, “CallRateBelowThreshold”, “Off-target variant (OTV)”, “Hemizygous”, or “MonoHighResolution” (or had previously failed Affymetrix’ internal validation process) were removed. A smaller number of DNA samples derived from saliva were also genotyped and clustered separately but are excluded here. Prior to statistical analysis, samples with sex discrepancies (reported versus expected), unexpected relatedness (half-sib or greater), and replicate samples with the lowest call rate were excluded. All statistical analyses were performed in [R] [91].
