seXY: a tool for sex inference from genotype arrays

Motivation: Checking concordance between reported sex and genotype‐inferred sex is a crucial quality control measure in genome‐wide association studies (GWAS). However, limited insights exist regarding the true accuracy of software that infer sex from genotype array data. Results: We present seXY, a logistic regression model trained on both X chromosome heterozygosity and Y chromosome missingness, that consistently demonstrated >99.5% sex inference accuracy in cross‐validation for 889 males and 5,361 females enrolled in prostate cancer and ovarian cancer GWAS. Compared to PLINK, one of the most popular tools for sex inference in GWAS that assesses only X chromosome heterozygosity, seXY achieved marginally better male classification and 3% more accurate female classification. Availability and Implementation: https://github.com/Christopher‐Amos‐Lab/seXY Contact: Christopher.I.Amos@dartmouth.edu Supplementary information: Supplementary data are available at Bioinformatics online.