CuCodeML: GPU-Accelerated CodeML for the Branch-Site Model in Phylogenetics

CodeML is the key program in the software package PAML, which is widely used in phylogenetic analyses of DNA and protein sequences. CodeML is often used for detecting sites under positive selection, but is time-consuming when performing large-scale analyses. Due to their computational power, graphics processing units (GPUs) are used to accelerate software for a range of applications, which motivates using GPUs to speed up CodeML. In this paper, we present CuCodeML, a GPU version of CodeML for the branch-site model, which is a widely used approach for the analysis of positive selection. In our single-CPU server with two Nvidia Tesla K20 GPUs, experiments on large data sets show up to 76× speedup with 1 GPU (vs. CodeML run normally), and up to 129× speedup with 2 GPUs. CuCodeML is available from https://github.com/rmingming/cucodeml.