Alex Wade - Optimising Genomics Workflows for Neoverse - AHUG SC21

Abstract: This work represents the joint co-design efforts of Arm, Atos, the Leibniz Supercomputing Centre (LRZ) and UCL in the domain of genomics with a specific application to the prediction of mutations in Mycobacterium tuberculosis (TB) which confer drug resistance and how to treat patients on an individual basis.

We aim to extend work carried out at Oxford and UCL on the influence of sequence on binding affinities to investigate the resistance of Mycobacterium TB to rifampin. Rifampin is a critical component of the multi-drug regimens used to treat tuberculosis. Whilst development of resistance to rifampin is rare compared to other anti-TB drugs, it often occurs in conjunction with other resistances leading to higher instances of treatment failure and mortality. Understanding of which mutations confer resistance is key to drug susceptibility testing and improving patient outcomes. Rifampin resistance is mediated by mutations in the rpoB gene, which codes for the beta-subunit of RNA polymerase. Thus, our work will study the relative binding affinity changes associated with amino acid mutations in the rpoB protein.

To this end, the work now underway will first investigate the performance of gene sequencing codes such as viridian and racon on new hardware provided by Arm. This hardware will consist of one Ampere Altra E252-P30 2U Mt Snow ""Edge"" server with 80 cores based on the new Arm Neoverse-N1 CPU core design. This will be followed by additional studies into the scaling and performance of molecular dynamics codes such as GROMACS and NAMD. In these endeavors the expertise of both Atos and LRZ will be leveraged. Atos will provide expertise in the optimization of code to best utilize new architectures. LRZ will provide expertise in measuring the performance and scalability of these codes as well as providing access to bespoke software solutions to automate these measurements based on earlier work performed on their own testbed called BEAST which already features some earlier Arm hardware.

Combining expertise from all these sources will allow for optimization of both software and hardware in this co-design effort to meet the demands of impactful and cutting-edge genomics workflows. Workflows which are already deployed in clinical settings by collaborators at Oxford but now will be optimized to exploit Arm's newest architectures to their full potential.

Speaker bios: Dr. Wade is a post-doctoral researcher at UCL focused on running simulations on high performance computers. These simulations investigate human proteins with a view to testing the efficacy of drugs and designing new ones to treat disease.