BLINK is a Genome Wide Association Study (GWAS) Method (Giga Science, 2019), standing for Bayesian-information and Linkage-disequilibrium Iteratively Nested Keyway (BLINK). BLINK is an enhanced version of FarmCPU GWAS method (PLoS Genetics, 2016), which stands for “Fixed and random model Circulating Probability Unification”. Similar to Multi-loci Mixed Linear Model (MLMM), both BLINK and FarmCPU use multi-locus model for testing markers across genome. BLINK conducts two fixed effect models iteratively. One model tests marker one at time with multiple associated markers fitted as covariates to account for population stratification. The other model selects the covariate markers to directly control spurious association instead of kinship, unmasking the confounding between testing marker and kinship. BLINK eliminate the requirement that genes underlying a trait are distributed equally across genome to further improve statistical power. BLINK also replaces the REstricted Maximum Likelihood (REML) in a mixed linear model in FarmCPU with Bayesian Information Content (BIC) in a fixed effect model to boost computing speed.
BLINK method was implemented in both R and C languages. The R version was developed by Dr. Yao Zhao (Email: firstname.lastname@example.org) and can be accessed as stand-alone version on GitHub, or through GAPIT. The C version was developed by Dr. Meng Huang (Email: email@example.com). Bit operations were used in the C version. As a result, BLINK C version is not only faster than BLINK and FarmCPU R versions, but also faster than PLINK. The C package is available at GitHub, containing executable program.