FastQTLmapping: an ultra-fast package for mQTL-like analysis

Here we describe fastQTLmapping, a C++ package that is computationally efficient not only for mQTL-like analysis but as a generic solver also for conducting exhaustive linear regressions involving extraordinarily large numbers of dependent and explanatory variables allowing for covariates. Compared to the state-of-the-art MatrixEQTL, fastQTLmapping was an order of magnitude faster with much lower peak memory usage. In a large dataset consisting of 3,500 individuals, 8 million SNPs, 0.8 million CpGs and 20 covariates, fastQTLmapping completed the mQTL analysis in 7 hours with 230GB peak memory usage.