FaST linear mixed models for genome-wide association studies

We describe FaST-LMM, a linear mixed model for genome-wide association studies that scales linearly in the number of individuals in both runtime and memory use. Our algorithm is an order of magnitude faster than current efficient algorithms (EMMAX/P3D) on Wellcome Trust data with 15,000 individuals. On synthetic data, FaST-LMM can analyze 120,000 individuals in just a few hours, whereas the current algorithms are unable to analyze even 20,000 individuals (http://fastlmm.codeplex.com).