BLINK

BLINK performs genome-wide association studies by iteratively selecting markers and refining models using linkage disequilibrium and Bayesian information criteria to improve quantitative trait nucleotide (QTN) identification, statistical power, and computational efficiency.

Key Features:

• Iterative Model Optimization: Iteratively optimizes mixed linear models by alternating marker selection and model evaluation, replacing random effect models (REM) with fixed effect models (FEM) guided by Bayesian information criteria.
• Incorporation of Linkage Disequilibrium: Utilizes linkage disequilibrium (LD) information instead of assuming even QTN distribution to improve marker selection and localization of associated variants.
• Enhanced Computational Efficiency: Substantially reduces computation time for large GWAS datasets (reported example: ~3 hours for one million individuals and 500,000 markers versus up to a week with FarmCPU).

Scientific Applications:

• Complex trait mapping: Identification of QTNs and genes associated with complex human diseases and agriculturally important traits.
• Large-scale GWAS: Analysis of very large cohorts and high-density marker datasets where computational efficiency is critical.
• Comparative GWAS methods: Replacement or augmentation of FarmCPU-style approaches to increase statistical power and speed in association studies.

Methodology:

Iteratively refines models by integrating linkage disequilibrium data with Bayesian information criteria to select markers, replacing REMs with FEMs during iterative optimization of mixed linear models.