Most of the important properties of molecules are determined by the way they interact with other molecules, and this is particularly true of biomolecules that almost all rely on selective intermolecular interactions for their function. Despite advances in our understanding of intermolecular interactions at the level of simple functional group contacts, such as H-bonding, hydrophobic interaction, aromatic stacking etc, it is still an extremely challenging task to predict the structure of intermolecular complexes of even relatively simple small molecule complexes, let alone protein-protein interactions. The development of fast and accurate methods for computationally estimating the three-dimensional structures and thermodynamic stability of intermolecular complexes remains a major scientific challenge. The state-of-the-art is either all atom simulations, where the force-fields are still too unstable to allow accurate estimation of free energies and the calculations are too large to use in any kind of screening programme, or docking algorithms that use simple empirical scoring functions, which are fast but crude. The research programme proposed here aims somewhere in the middle. We will use methods that have a sound theoretical basis and are sufficiently robust to calculate accurate free energies, combined with a stripped down representation of molecular structure to allow rapid calculations on macromolecules and large compounds libraries. The ability to dock two molecules to predict the structure and stability of the intermolecular complex has obvious applications in all areas of biology and medicine, where protein-ligand, protein-protein, protein-DNA complexes regulate almost all biochemistry in the cell. This is a challenging target and there are many aspects to the problem for which no current solutions exist, eg handling of large scale loop movements in proteins that alter the nature of interaction interfaces, but we will start with simpler systems and gradually work towards these more difficult problems. The academic applicant has developed a computational approach for small molecules that accurately estimates the stability of intermolecular complexes in solution based on a reduced representation of the electrostatic fields of binding partners. The industry partner has developed a computational approach for small molecules that accurately describes the structural properties of intermolecular complexes based on a different representation of the electrostatic field. The aim of this project is to develop a new composite computational method that takes key elements of these two approaches and combines them for the accurate prediction of the structure and stability of biomolecular complexes in solution. This will establish a new tool that could be applied to a wide range of problems in biology. Discovering what small molecules or macromolecules are likely to bind to a specific target protein would be of immense value in progressing from the information contained in genomes to an understanding of the functional biochemistry of living organisms. The approach outlined here provides a promising strategy for improving our chances of success in this endeavour.

Advances in High Performance Computing (HPC) and scientific software development will have increasingly significant societal impact through the computational design of new products, medicines, materials and industrial processes. However, the complexity of modern HPC hardware means that scientific software development now requires teams of scientists and programmers to work together, with different and non-overlapping skill-sets required from each member of the group. This complexity can lead to software development projects stalling. Investments in software development are in danger of being lost, either because key members of a team move on, or because a lack of planning or engagement means that a sustainable user and developer community has failed to gel around a particular code. Research Software Engineers (RSEs) can solve this problem. RSEs have the skills and training necessary to support software development projects as they move through different stages of the academic software lifecycle. Academic software evolves along this lifecycle, from being a code used by an initial team of researchers, through to a large multi-site community code used by academics and industrialists from across the UK and around the World. RSEs provide the training and support needed to help academic software developers structure their projects to support the sustainable growth of their user and developer communities. RSEs are also highly skilled programmers who can train software developers in advanced HPC techniques, and who can support developers in the implementation, optimisation and testing of complex and intricate code. Together with academic software developers, RSEs can support UK investment in HPC, and ensure that the potential of computational science and engineering to revolutionise the design of future products and industrial processes is realised. This project aims to develop sustainable RSE career pathways and funding at Bristol. This will support the growth of a sustainable team of RSEs at the University. Software development projects that will be supported include; the building of code to interface real biological cells with virtual simulated cells, so to support the rapid design of new biomanufacturing control processes; the development of code to more quickly model the behaviour of electrons in novel materials, to support the design of new fuel cells and batteries; code to improve our understanding of glass-like matter, so to help design new materials with exciting new properties; software to support modelling of the quantum interaction between laser light and microscopic nanoparticles, to support the design of optical tweezers and new optically driven nanomachines; and code to design new medicinal drugs and to understand why existing treatments are no longer working, thereby supporting the development of 21st century medicine. Finally, this project aims to create a coherent set of teaching materials in programming and research software engineering. These, together with the development of software to support science and programming lessons held in an interactive 3D planetarium, will help inspire and educate the next generation of scientists and RSEs. These materials will showcase how maths, physics, computing and chemistry can be used in the "real world" to create the high-tech tools and industries of the future.

Nobel Laureate Richard Feynman in his Lectures on Physics famously remarked that "...everything that living things do can be understood in terms of the jigglings and wigglings of atoms". This deceptively simple statement highlights the difficulty that structural biologists, medicinal chemists and computational scientists are faced with when attempting to understand human health and disease. We are used to thinking about a static, isolated picture of objects at the atomic scale, but often it is the dynamics (the "jigglings and wigglings") of the system and its environmental interactions that determine the underlying science, such as the role of intrinsically disordered proteins in neurodegenerative diseases or the possible link between quantum entanglement and molecular vibrations in biological photosynthesis. Twentieth century science not only set the challenge of studying life at the level of the structure and dynamics of atoms, but also provided (in theory) the solution, through the laws of quantum mechanics and the famous Schroedinger equation. Quantum mechanics explains the fundamental behaviour of matter at the atomic scale, and smaller. It enables scientists to make predictions about materials that are inaccessible to experiment, such as the structure of solid hydrogen in a star's core. At a more everyday level, quantum mechanics is routinely used by researchers in the microelectronics and renewable energy industries to rapidly scan multitudes of hypothetical materials compositions. In this way, the costly manufacturing process of the new materials need only begin once the desired properties have been predicted. However, quantum mechanics does not directly enable scientists to understand the biomolecular origins of disease, or to design new medicines to combat it. The reason for this comes down to Feynman's statement. It is infeasible to solve (even approximate) equations of quantum mechanics for the length and time scales sufficient to model all of the atomistic movements that need to take place, for example, for a drug molecule to find its target. Instead, computational chemists use a much simplified computational model, known as a force field, to estimate the dynamics of atoms. The force field models the atoms as bonded together in a molecule by springs, and interacting with other atoms through electrostatic and van der Waals forces, which are much stronger than gravity at the atomic scale. The strengths of these interactions are modelled by thousands of adjustable parameters, which have been manually tuned to reproduce experimental data over a period of many decades. We are reaching a stagnation point where accuracy is urgently needed for computer-aided design of new medicines, but parameter tuning delivers only small improvements. My vision for this UKRI Future Leaders Fellowship is to build a multi-disciplinary team that will work together to close the accuracy gap between quantum mechanics, and the approximate force fields used in biology and medicine. By working with international coding efforts, I will build the theory and software infrastructure required to dispense with these adjustable force field parameters, and instead derive them directly for the system under study, such as a protein implicated in disease. This will enable me to build more accurate computational models of the electrostatic and van der Waals interactions that determine the strength of binding of potential drugs to their targets. By crossing disciplinary boundaries to train in data science and machine learning, I will deploy the expertise that has been made famous for its applications in face and speech recognition, to create a spectrum of tools for speeding up the assignment of parameters and improving the accuracy of force field design. Finally, by undertaking secondments in the pharmaceutical industry, I will ensure that the developed methods will be used for the cost efficient design of the next generation of medicines.

Atomistic simulations are the main application of high-performance computing research, and increasingly underpin innovative R&D processes in the chemical and life sciences industry. OpenMM is the fastest growing atomistic simulation engine among the current ecosystem of open-source academic software. Originally targeting a biomolecular simulation audience, the OpenMM user base is growing exponentially and has permeated diverse related domains, including materials modelling, quantum chemistry, structural bioinformatics, chemoinformatics, artificial intelligence and machine learning. The success of OpenMM is down to a design that achieves an excellent tradeoff between extensibility (via a robust user interface) and performance on GPUs (via auto generated CUDA kernels) for molecular dynamics (MD) simulations. OpenMM is used standalone or via plugins to other atomistic simulation engines, providing access to GPU-accelerated MD simulation capabilities for the whole atomistic simulation ecosystem. We have surveyed the OpenMM user community to identify its most pressing needs. OpenMM is currently maintained by a single core developer who can no longer support the training and support needs of its rapidly growing user community. We will transition OpenMM to a more sustainable community-driven development model. We will develop training resources to upskill users, and engage the community to widen participation in developing and maintaining OpenMM functionality. Machine learning (ML) potentials have the potential to revolutionise the future of atomistic simulation methodologies. Our community survey has identified strong interest in ANI neural network and GAP Gaussian process regression methods. We will deliver a self-contained GPU-optimised GAP implementation in OpenMM and coordinate with project partners working on an OpenMM ANI implementation to offer the community a library of ML potentials that can be readily plugged into existing simulation engines. OpenMM must adapt to scientific (ML potentials) and technological (increased hardware heterogeneity) drivers to continue offering its user base an optimised tradeoff between speed and ease of modification over the coming decade. We will integrate in OpenMM a multiple level intermediate representation compiler (MLIR) to auto generate from user-specified Python instructions optimised low-level code targeting diverse hardware. By enabling users to specify custom atomic featurisation techniques as OpenMM operations, which can be finely interleaved with Tensorflow or Pytorch operations, we will position OpenMM as the simulation engine of choice to support deployment of next generation ML potentials onto current GPUs and emerging AI-hardware accelerators. Our community has also required support to facilitate the combined use of independently developed OpenMM software solutions with other software from the broader atomistic simulation ecosystem. This research will develop a standardised interface to integrate OpenMM community software with CCPBioSim's interoperable Python framework BioSimSpace. We will demonstrate integration of all the work packages of this research via production of GAP ML pipelines for two use cases that target grand challenges in soft-condensed matter modeling (organocatalysis - recently recognised by the 2021 Nobel Prize in Chemistry- and protein-ligand binding). Altogether this research will position the OpenMM user community at the forefront of next-generation hybrid machine learning/molecular mechanics potentials for soft-condensed matter modelling. Deeper integrations with AI and HPC communities will pave the way for atomistic simulations to harness emerging exascale opportunities. Transitioning from a single developer to a community-driven development governance model will improve sustainability of the codebase and encourage greater adoption of OpenMM in associated academic communities and industry.

Biomolecular simulation is a fast growing area, making increasingly important contributions to structural biology and pharmaceutical research. Simulations contribute to drug development (e.g. in structure-based drug design and predictions of metabolism), design of biomimetic catalysts, and in understanding the molecular basis of disease and drug resistance. CCP-BioSim (ccpbiosim.ac.uk) was established in 2011 with support from EPSRC to strengthen molecular simulations at the life/sciences interface, and develop links with academia/industry. CCP-BioSim led in 2013 a successful EPSRC bid for a High-end Computing consortium in Biomolecular simulation, HECBioSim (hecbiosim.ac.uk). HEC-BioSim works to bring high-performance computing to a wider community of experimentalists and to engage physical scientists in biological applications. CCP-/HECBioSim regularly organize training workshops and provide a framework for networking and collaboration. We also work to develop and apply advanced methods, and engage with international activities (e.g. NSF, CECAM, NIH etc.). We actively engage with structural and chemical biologists and industrial researchers through collaboration, dissemination and application of software, and invitations to conferences and workshops. We actively collaborate with other CCPs via joint workshops and conferences. We actively support community software development and have released software to make biomolecular simulations more accessible to diverse communities. Our field benefits from continuous advances in HPC and chemical physics (e.g. multiscale modelling, 2013 Nobel Prize in Chemistry). Our techniques have reached a stage where we now aim to comprehensively transform the science of molecular design. Pharmaceutical companies continuously seek to design new drugs to treat e.g. antibacterial infections or cancers. Agrochemical companies continuously seek new chemicals to treat pests, supporting agricultural growth to secure food for our population. Biomolecular design is a complex multi-objective optimization problem. To make significant headways our field is increasingly combining multiple software packages into workflows. This departs from the historical paradigm of our field, where research problems were tackled with one or a few techniques at a time. Our community lacks software to easily assemble our tools into robust, scalable and comprehensive workflows needed to address the science of molecular design. As a CCP-/HECBioSim flagship community software project, we propose to develop BioSimSpace. Our software will provide an interoperability layer to allow software packages from our communities to work together. Translation tools will ensure that outputs from one package can be easily used as inputs to another package. Importantly, BioSimSpace will enable components of a workflow to be written such that are independent of the underlying software application. This will allow workflow components to be mixed and matched into more complex workflows, and for those workflows to select applications that will be optimal for the underlying computer hardware. We will use BioSimSpace to validate new workflows that address the grand challenges of screening drugs for potency, binding pathways and kinetics. By working with a commercial software vendor, we will make it easy to package BioSimSpace-based components so that they can be easily shared, installed and sold via a software marketplace. By working with a range of national and international industrial and academic partners, we will develop and apply BioSimSpace-based workflows to address molecular design problems faced by our community, and the pharmaceutical and agrochemical industries. By using supercomputers we will demonstrate how large BioSimSpace workflows help decrease the costs and time needed to design molecules for healthcare and industrial biotechnology applications.