Portability and scalability of MD codes for alchemical binding free energy calculations

Binding free energy calculations allow for the affinity of a drug to a protein to be quantified. This affinity for many inhibitors can be used as a proxy for their effectiveness as drugs to treat a disease. Binding free energy calculations can then be used as a key tool in drug design and personalised medicine endeavours. These calculations can be used to assess changes to a drug that could improve its effectiveness or find which drug would be most effective for an individual patient taking into account their unique genome. For alchemical binding free energy calculations to become ubiquitous in personalised medicine and drug discovery the molecular dynamics codes they use must be reliable but also more portable and scalable. Moreover, the alchemical methods, which currently require significant user expertise, must become more usable and accessible. To meet these demands, we have developed a pipeline for the preparation, execution and analysis of these calculations on heterogeneous HPC environments called TIES. This talk will outline the TIES method focusing on its application to varied HPC resources at scale. This talk will also cover our findings regarding the investigation of Intel's next generation solution for software portability, namely OneAPI.