SL_REML

SL_REML is a software tool that introduces two novel algorithms, SLDF_REML and L_FOMC_REML, for efficient estimation of variance components in linear mixed-effects models (LMM) used in genome-wide association studies (GWAS). These algorithms address the computational burden of residual maximum likelihood (REML) estimation by leveraging the principle of Krylov subspace shift-invariance, allowing for a single round of iterative matrix operations, followed by repeated objective evaluations using vector operations, resulting in faster computation compared to existing methods. SLDF_REML can also utilize precomputed genomic relatedness matrices to enhance computational efficiency further.

Numerical experiments demonstrate that SL_REML's interpreted-language implementations of these algorithms match or surpass the speed, accuracy, and flexibility of existing compiled-language software packages, making them suitable for integration into current GWAS LMM software.