BioExcel-2 deliverable D3.6 – Pre-Exascale showcase calculation and Use Case Final Report

This deliverable is an update to the project half-time deliverable D3.3 - Use Case Progress Report. It provides a final report on BioExcel-2’s demonstrator research projects (the “Use Cases”), including a pre-exascale showcase calculation performed during the second half of the project. Final reports for each Use Case include a summary of work done, scientific results obtained, software and HPC resources used, and how the work has impacted the community. We demonstrate that the Use Cases serve as exemplars of how important challenges in biomolecular modelling and simulation can be tackled effectively using HPC resources and various combinations of the software developed and supported by the CoE. The Use Cases are at the core of demonstrating the impact of BioExcel's science-enabling software development to the benefit of the biomolecular modelling and simulation community. As well as providing valuable scientific results in their own right, they are a rich source of expertise and best practice methodologies which are drawn upon by the Centre to provide support and training to the community and to engage with industry.

The pre-exascale showcase calculation (Use Case 5 - UC5), demonstrates the readiness of BioExcel applications GROMACS and PMX to enable state-of-the-art methods to obtain results of equivalent accuracy to leading commercial software, and to do so with unprecedentedly fast turnaround time by using preexascale HPC resources. The simulations involved were performed over a period of 72 hours using 480 nodes / 50k cores - 93% of a 3.5 Pflop/s (peak) supercomputer (“Raven” hosted at MPCDF). This demonstrates the potential for large-scale rapid-turnaround virtual drug screening that would currently take weeks for an individual researcher to perform based on available compute time but which could become routine in the exascale era, with significant potential to massively accelerate drug discovery.

Use Cases 1 and 3 have demonstrated that computational pipeline protocols employing various combinations of the core BioExcel applications GROMACS, HADDOCK and PMX can be used for mutational analysis and to perform free energy and docking calculations for the design of antibody-based and small ligand therapeutics. The BioExcel-developed Building Blocks (BioBB) library for interoperable biomolecular simulation workflows in conjunction with the PyCOMPSs task-based execution manager has enabled many of these pipelines to execute efficiently on HPC resources, demonstrating how the ultimate goal of shortening the time and effort to develop new and better therapeutics may be achieved.

Use Case 2 has shown how parallel execution of containerised HADDOCK and the necessary data movement underpinning can be orchestrated by PyCOMPSs, demonstrating how (pre-)exascale HPC resources will be able to be used to study significant fractions of the interactions of biomolecules encoded by the human genome.

Use Cases 4a and 4b have demonstrated how QM/MM simulation with GROMACS, CP2K and CPMD using HPC resources to perform costly quantum calculations are set to improve electrospray ionization mass spectrometry - a key analytical technique in proteomics - and, combined with free energy calculations using PMX, are enabling the design of fluorescent proteins enabling the high-resolution monitoring of cellular functions, gene expression, protein-protein interactions, intra-cellular interactions in living systems as well as understanding and finding novel strategies to tackle disease