Svishcheva, G.R. et al. Nat. Genet. 44, 1166–1170 (2012).

Attempts to associate human disease with underlying genetic variation have been buoyed by a rising tide of genotyping data. At the same time, powerful methods are needed to tease out real associations from those due to the cryptic relationships and shared ancestry found in large-sample studies. Mixed models can remove these confounding effects but are computationally demanding, thus leading to recent efforts to improve their efficiency. Svishcheva et al. introduce GRAMMAR-Gamma, which is based on a fast approximation of the powerful FASTA method. The approach renders computational time linear rather than quadratic with respect to the number of individuals. Tests on synthetic and real data demonstrate that millions of nucleotide variants can be rapidly analyzed from large-sample data sets.