Overview / Usage

High-order epistasis detection is a bioinformatics application with very high computational complexity that searches for correlations between genetic markers such as single-nucleotide polymorphisms (SNPs), single base changes in a DNA sequence that occur in at least 1% of a population, and phenotype (e.g. disease state). Finding new associations between genotype and phenotype can contribute to improved preventive care, personalized treatments and to the development of better drugs for more conditions.

Showcased during the OneAPI Dev Summit 2020 keynote: Joe Curley, with guest appearance of Aleksandar Ilic. “oneAPI Vision for Heterogeneous Compute”, Intel oneAPI Developer Summit, keynote, virtual event, November 2020

Methodology / Approach

Parallel architectures are being targeted to efficiently perform the highly data-parallel computations that implement epistasis detection searches.

The Intel DPC++ Compatibility Tool has been used to seamlessly port a CUDA-based implementation which is related to the following publication. More than 95% of our hand-tuned code got automatically migrated.

R. Nobre, S. Santander-Jiménez, L. Sousa and A. Ilic. Accelerating 3-way Epistasis Detection with CPU+GPU processing. Paper presented in 23rd Workshop on Job Scheduling Strategies for Parallel Processing (JSSPP), New Orleans, 2020. DOI: 10.1007/978-3-030-63171-0_6

Technologies Used

This application has been obtained through translation of an OpenMP+CUDA code using the Intel DPC++ Compatibility Tool.

DevCloud nodes with Intel Iris Xe Max GPUs (Gen12) and state-of-the-art Intel Xeon CPUs are used.

Documents and Presentations

Repository

https://github.com/rjfnobre/crossarch-episdet

Collaborators