Improving The Longevity Of New Infectious Disease Therapeutics Using Machine Learning / Artificial Intelligence In Early Stage Drug Discovery

Drugs against diseases caused by viruses, bacteria and parasites have transformed human health and saved millions of lives. Nevertheless, their widespread use and misuse has led to the emergence of antimicrobial resistance (AMR) that poses a potentially catastrophic threat to public health. The increasing power of genomic sequencing is offering new ways to rapidly detect and respond to the development of antimicrobial resistance. The availability of this wealth of data, along with the latest developments in artificial intelligence / machine learning (AI/ML) techniques, allows the development of sophisticated approaches that can fully leverage this data to pre-empt the effects of potential resistance mutations.

The aim of this project is to develop new computational tools for automatically analysing the molecular consequences of single nucleotide polymorphisms (SNPs) linked with therapeutic resistance from genome wide association studies (GWAS) and to use this wide-ranging iterative analysis to build a predictive model to identify future SNPs that could lead to therapeutic failure. It will focus on analysing point mutations, singularly and as observed haplotype combinations, that represent one of the major routes to resistance. It will leverage the wealth of in-house and publicly GWAS linking SNPs to drug resistance. By exploiting the latest state-of-the art tools for predicting various measures of the effect of a mutation, this offers an exciting opportunity to measure the biophysical functional, geometrical and genomic effects these mutations are having on proteins on a large scale across different organisms and resistance types.

Such insights will be used to develop a novel computational tool utilising the latest developments in machine leaning to anticipate mutations leading to resistance before they become fixed in a given pathogen population. The ability to effectively predict the effects of such mutations has several applications including in the development of the next generation of drugs. The tool can be used to make an informed decision as to the effects of mutations on a potential drug binding region, as well as connected distal regions, that might lead to the drug becoming less effective. If areas that are tolerant of mutations can be avoided, drugs with a longer clinical life can be developed. This is important in pathogen drug management, and especially for neglected tropical diseases, where new therapeutics are difficult to develop and market incentives for developing new drugs are weak.

These new tools will be validated building on our preliminary research in tuberculosis by independently predicting likely resistance mutations in target proteins from studies currently being undertaken by us. As well as helping validate our method it will also provide new insights into routes of resistance. Moreover, the new tools will be applied by us and collaborators within our existing drug discovery programs against a range of infectious disease agents e.g. schistosomiasis. The results of the application to real situations along with validation will be used to iteratively feedback and improve the machine learning tool.

This research project takes a novel approach to addressing the critical threat of AMR as highlighted in the 2014 UK Review on Antimicrobial Resistance. The new approaches can be used to inform the development of novel therapeutics and in turn promises to answer specific questions around the modes of resistance in individual infectious disease organisms. It has the future potential to be applied in identifying early emergence of resistance and in clinical diagnostics. By contributing new tools and methods it will help ease the burden of AMR, before it becomes a much larger drain on our healthcare system.

Antimicrobial resistance threatens the effective prevention and treatment of an ever-increasing range of infections globally. The increasing power of sequencing is offering new ways to rapidly detect and respond to the development of antimicrobial resistance. To fully leverage this information, sophisticated approaches are required to identify potential resistance mutations. This project exploits the wealth of in-house and publicly available genome wide association studies linking single nucleotide polymorphisms to drug resistance. It utilises the latest tools for measuring the biophysical effects of these mutations along with measures of the effects on protein function, experimental phenotype and genomic measures of resistance. This will be used to construct new tools exploiting the latest developments in machine learning to anticipate mutations leading to resistance before they become fixed in a population. It will build and extend the successful preliminary implementation of data collection, curation and modelling and subsequent analysis made by us in the study of resistance to first and second line treatments against Tuberculosis. This new computational method will be applied to significant human pathogens. Firstly, through validation based on new data from studies of M. tuberculosis, with the additional outcome of potentially discovering new potential mutations that could lead to resistance. Secondly, the application to existing drug discovery pathways led by us in the development of novel therapeutics to treat schistosomiasis and within programs managed by collaborators targeting a wider range of infectious diseases. The results of the application to real situations along with validation will be used to iteratively feedback and improve the machine learning tool. The new approaches can be used to inform the development of novel therapeutics and has the potential to be applied in identifying early emergence of resistance and in clinical diagnostics.

The emergence of resistance to drugs against pathogens (viruses, bacteria and parasites), termed antimicrobial resistance (AMR), poses a catastrophic threat to public health. By developing novel and innovative tools using the latest in artificial intelligence / machine learning (AI/ML) techniques to anticipate the effects of mutations leading to resistance before they become fixed in a given pathogen population, this project will have great impact in the development of a new generation of pathogen therapeutics. Through understanding the molecular consequences of mutations associated with resistance, this research will effectively contribute to extending the lifetime of new infectious disease therapeutics. Improving the longevity of antimicrobial chemotherapies will help reduce the costs of drug delivery for healthcare systems and improve the outcomes from the millions of people globally affected by infectious disease.

Some of the immediate key beneficiaries of the project outcomes are research and development scientists working to produce the next generation of antimicrobial therapeutics, especially those employing a target based approach. They will gain from the insights gained from understanding the molecular consequences of mutations associated with antimicrobial resistance that can be used to improve and generate new therapeutics. The new computational methods developed within this project will be able to anticipate novel mutations, yet to be fixed within a population, that might impact on therapeutic efficacy and will thus be complementary to other state-of-the-art techniques in pathogen surveillance and mutation analysis. Moreover, improving the knowledge of the molecular consequences of genetic variation associated with resistance will be of great interest to scientists investigating the surveillance of antimicrobial resistance using genomic or other techniques and those working at the clinical interface in the development of diagnostics to a wide range of pathogens.

The focus on applying and validating the computational methods within existing drug discovery pathways being pursued within the Furnham group and with national and international collaborators will mean the findings from this project will have direct impact to the development of new drugs against a range of infectious diseases. In addition, using important pathogens such as M. tuberculosis that exhibit significant resistance to a range of current treatments for validating the tools will impact on our understanding of drug resistance in these pathogens.

The proposed research is highly interdisciplinary and will allow those involved in the project to receive not only transferable technical skills and knowledge, but also to form a network of contacts and working relationships amongst the academic and industrial communities. The staff employed on the project will benefit from continued development of their scientific skills, further building knowledge and expertise in AMR and AI/ML research. The application of machine learning contributes to the UK's Industrial Strategy, advancing innovation and developing UK capacity in this technology area.

Furthermore, the project outcomes have the potential to be applied to any antimicrobial target, which will be of great interest to those working on a range of pathogens in both human health and animal health. This contributes to both One Health concept and AMR priority areas set by UKRI. It will advance the cross-council initiative to tackling AMR by addressing challenges in two themes: understanding resistant bacteria and accelerating therapeutics and diagnostics development. The project also helps in the world-wide effect to combat AMR brought together under the initiative of the WHO.