A supercomputing tool to deliver rapid drug discovery
The emergence of digital tools involving high-performance computing (HPC) and big data modelling has opened up drug design to non-traditional health industry players such as big tech companies who are challenging the notion that biological systems are too complex to model. “We are witnessing a shift from models based on data, to data being generated because it is suitable for training models,” notes Andrea Beccari, head of discovery platforms at Dompé farmaceutici, the project host. LIGATE’s main goal was to affirm the role supercomputing resources can play in drug discovery, especially crucial in times of acute health crises such as pandemics. The project developed an artificial intelligence (AI) assisted virtual drug screening tool that is adaptable to any hardware architecture, and forms the functional core of the Exscalate drug discovery platform. The Exscalate platform enables researchers to test a library of molecules against a range of targets, to identify possible treatments for threats such as COVID-19. “Initially, trillions of virtual molecules will be screened to find suitable broad-spectrum antiviral drugs. The best candidates will be tested with antiviral assays and structural analysis methods,” says Beccari. The LIGATE tool generates a 3D simulation of molecules of interest, which are used to explore all the molecular conformations (atomic arrangements) possible within the active site of the target protein. The tool selects the most promising based on the molecule’s interactions with the protein’s amino acids, and an AI tool ranks each based on their predicted efficacy.
Portability and flexibility
The LIGATE solution integrates both open-source and proprietary components. Open-source modules were domain-agnostic to support the entire application’s development and large-scale deployment, and domain-specific for the molecular simulation and virtual screening functions. “LIGATE benefited from the flexibility, efficiency and scalability of community-driven code, complemented by proprietary modules which allow for its commercial exploitation,” adds Beccari. To take advantage of forthcoming exascale computing resources, compatibility with various hardware architectures, especially GPUs, was crucial. This is particularly true for urgent computing applications – those reliant on the maximum computing power (often using multiple supercomputers simultaneously) to respond to critical threats, such as the COVID-19 pandemic. But as much of the key hardware is written in proprietary programming languages, the team had to develop a programming model (using royalty-free SYCL) of the main modules, making the software portable across CPUs, GPUs and FPGA cards.
Increasing the accuracy of drug evaluation
A key project aim was to implement the screening tool within the EU-supported LEXIS platform, established to provide greater access to powerful HPC resources. Using a web interface designed to make it even more accessible, the LIGATE tool was demonstrated at the last EuroHPC Summit, when it ran simultaneously on the Karolina, LUMI and LEONARDO supercomputers. “Researchers will soon be able to use a network of European computational resources to discover new drugs without having advanced computing expertise,” explains Beccari, who acted as LIGATE project coordinator. A predecessor project, EXSCALATE4CoV, used an earlier version of the LIGATE tool to study repurposed drugs for suitability against COVID-19, identifying Raloxifene as a promising treatment. The drug subsequently passed a phase II clinical trial. “Remarkably, during molecular simulations, the platform also found a potentially safer new vaccine, now being preclinically evaluated,” adds Beccari. The project was carried out with support from the European High Performance Computing Joint Undertaking (EuroHPC JU), an initiative set up to develop a world-class supercomputing ecosystem in Europe.
Keywords
LIGATE, EuroHPC JU, molecule, drug, AI, drug discovery, virtual screening, supercomputers, portability, big data, GPU, virus, COVID-19, open-source, HPC