Phylogenetic reconstruction is one of the most important problems in biology to study the evolutionary relationships among organisms on Earth. It is vital inferences across the biomedical spectrum and “nothing in biology makes sense except in the light of evolution” (T. Dobzhansky).
IQ-TREE is a leading-edge and widely used software in evolutionary biology to reconstruct phylogenetic trees from genome-scale data, with >5,000 citations for the main paper published in 2015. Recently, IQ-TREE was instrumental in understanding the novel coronavirus (COVID-19) evolution and transmission.
IQ-TREE supports high-performance computing via shared (OpenMP) and distributed memory (MPI) parallelisation. However, usage of accelerators such as Graphics Processing Units (GPUs) is currently not supported in IQ-TREE. Offloading IQ-TREE's most computationally intensive kernels to GPU is expected to yield a notable performance boost, thereby enabling researchers to reconstruct more complex trees within record execution times.
This project aims to
(i) Accelerate phylogenetic reconstruction by using OpenMP to offload onto GPU IQ-TREE's key computational bottlenecks.
(ii) Benchmark the ensuing performance benefits on the Gadi GPU nodes at National Computational Infrastructure.
This project will greatly benefit the scientific community to deal with exponentially increasing amounts of genome-scale data and improve the global inference of phylogenies.
Programming models: C/C++, OpenMP.
Bioinformatics, HPC, GPU, algorithms