FarmCPU: An R Package of Genome-Wide Association Study for Big Data

Because of the open source packages and cross-platform support, R language is becoming popular in bioinformatics. A set of R packages is widely used in Genome-Wide Association Study (GWAS), such as GAPIT, GenABEL, MLMM, EMMA, among others. As the development of high throughput sequencing technologies have dramatically reduced the cost of genotyping, big data is coming to GWAS. We developed an R package named “FarmCPU” which is an acronym of “Fixed and random model Circuitous Probability Unification”. FarmCPU reads the data directly from the hard disk and saves lots of memory; FarmCPU iteratively uses a fixed model to scan the markers and a random model to build the fixed model and the time complexity of FarmCPU is close to a fixed model. To further increase the speed, users can also perform FarmCPU using multi CPUs. FarmCPU can deal with a data with 10,000 samples and 100,000,000 markers in a few hours using a laptop with single processor, which is currently difficult with the existing soft wares.