We are recruiting to a fully funded 4-year industry-track studentship in collaboration between the CDT in Biomedical AI and Exscientia to start in September 2021.
Modern drug design is a time consuming and costly process that often takes up to 10 years from identifying a drug target to, finding a drug candidate, and getting this candidate approved for clinical use. One way of speeding up this process is in the early stages of drug discovery, where we can make use of computational methods to come up with new drug candidate molecules and computationally predict how well they may bind and inhibit the function of the target protein. Only the most promising molecules are then synthesised reducing time and cost spent on synthesis. As a result, having reliable and accurate methods that can predict how well a drug-like molecule will bind to a target protein at a large scale (more than 1 million molecules) is crucial in speeding up the overall drug discovery process.
Many different approaches have been taken to address this problem from different docking and scoring approaches to dynamics-based approaches such as MMPBSA or alchemical free energy methods. It is broadly acknowledged, that alchemical free energy-based methods are most accurate, with binding affinity prediction accuracy in the range of 1 kcal/mol. Yet they are very computationally costly and can therefore only be used on a small subset of potential molecules. While docking and scoring methods can be used at a large scale, but often the docking score shows little to no correlation to experimentally measured binding affinities. One reason for these shortcomings can be attributed to the scoring function in terms of docking and the force field in terms of the alchemical free energy methods. The project will explore ways in which machine learning and more accurate quantum mechanics-based calculations can be leveraged to improve large scale binding affinity predictions.
Recent advances in machine learning meant that models can be used to learn potential energies from large datasets of density functional theory calculations of drug-like molecules. This means quantum level accuracy for predicting interaction energies from such machine-learnt potentials can be reached at a computational relevant for pharmaceutical timescales. There has been an initial proof of concept studies to show how these methods can be used to improve accuracies of alchemical free energy calculations with errors in comparison to experiment being as low as 0.5 kcal/mol. The idea behind this project is to explore the utility of these methods in the context of different affinity prediction strategies such as ensemble-based docking, as well as MD based methods. The goal will be to reach high accuracies at drastically reduced computational cost, providing a fast and accurate computational affinity prediction method in the early stages of drug design.
Applicant Profile
Applicants need to have a UK 2.1 honours degree, or its international equivalent, in a science discipline and interested in using mathematical, machine learning-based, and computational tools to address biomolecular problems.
CDT programme
Student appointed to this project will be part of the UKRI Centre for Doctoral Training in Biomedical AI. The CDT will train a new generation of interdisciplinary scientists who will shape the development of AI within biomedical research over the next decades. Our students will gain the skills, knowledge and acumen to realise biomedical breakthroughs using AI while anticipating and addressing the social issues connected with their research.
The CDT programme follows 1+3 format. In Year 1 you will study towards a Master by Research, undertaking a number of taught courses and taster research projects to broaden and refine your skills and explore different research areas. Following that you will begin your PhD project ML-Score: Exploiting Machine-Learnt Potential Functions for Drug Design co-supervised between the University of Edinburgh and Exscientia.