The University of Southampton
University of Southampton Institutional Repository

AI3SD Project: Predicting the Activity of Drug Candidates where there is No Target

AI3SD Project: Predicting the Activity of Drug Candidates where there is No Target
AI3SD Project: Predicting the Activity of Drug Candidates where there is No Target
This project aims to harness artificial intelligence and machine learning approaches to improve the discovery of new medicines. One of the most common situations in drug discovery is to know a molecule that possesses a desirable property and yet not to know how the molecule achieves that effect. The molecule needs improvement (often for things like how well it does its job, or how soluble it is in water) for it to be a realistic drug candidate and we must consider what changes need to be made. In this AI3SD project we know of molecules that efficiently kill the malaria parasite. The molecules need to be improved via small changes to their structures, yet we do not know the molecular biological target of these molecules. Without knowledge of the target it is impossible to design the improvements rationally. In such a project (known as “phenotypic drug discovery”) it is typically the case that the scientist will apply rules of thumb and intuition acquired over many related projects, in order to alter the structure of the molecule in search of those improvements. Yet it is also frequently the case that at such a stage of the project changes can be made to the molecule that accidentally obliterate the desired properties; most typically the molecules lose their potency against the pathogen. We will then have wasted time and resources making inactive molecules. On average each “fail” costs about two thousand pounds to make. It would be much more efficient if we could become more accurately predictive about which molecules need to be made.

In our research consortium, Open Source Malaria (OSM), we have twice tried and failed to generate predictive models. There have occurred, in the period since, major new advances in AI and ML, particularly in the private sector. Since all of OSM’s data and ideas are freely shared in the public domain, it is possible for us to work with anyone in the generation of new models. We have therefore in this project used the AI3SD funds to run a predictive modelling competition and elicited contributions from amateurs and leading AI companies. Several models were better than the others and so, in a crucial part of this project, we asked those winners to predict new molecules, molecules that have never existed before, that they predict will be effective at killing the malaria parasite. We then went to the lab to validate these predictions by making the molecules suggested, and measuring how effective they are at killing the parasite in blood. The result of this was that three of the six predictions were active, a “hit rate” of about the same as the human hit rate across the rest of the project. Interestingly, these actives included a couple of molecules that the human chemists would probably not have tried.

The end result of this work, aside from a new and predictive approach to the synthesis of antimalarial drug candidates, is be a case study of the actual capabilities of new AI/ML technologies in drug discovery: what works, what does not and an examination of why. Notably the project is still ongoing: since all the data and details of the approaches taken are in the public domain, others can try out their own predictive algorithms to see if they can do better.
AI3SD, Funded Project, Artificial Intelligent, Machine Learning, Drug Discovery
3
University of Southampton
Todd, Matthew
1084a7bc-bd6f-4014-8acd-091512a93a3f
Kanza, Samantha
b73bcf34-3ff8-4691-bd09-aa657dcff420
Frey, Jeremy G.
ba60c559-c4af-44f1-87e6-ce69819bf23f
Todd, Matthew
1084a7bc-bd6f-4014-8acd-091512a93a3f
Kanza, Samantha
b73bcf34-3ff8-4691-bd09-aa657dcff420
Frey, Jeremy G.
ba60c559-c4af-44f1-87e6-ce69819bf23f

Todd, Matthew , Kanza, Samantha and Frey, Jeremy G. (eds.) (2020) AI3SD Project: Predicting the Activity of Drug Candidates where there is No Target (AI3SD-Project-Series, 3) University of Southampton 14pp. (doi:10.5258/SOTON/P0042).

Record type: Monograph (Project Report)

Abstract

This project aims to harness artificial intelligence and machine learning approaches to improve the discovery of new medicines. One of the most common situations in drug discovery is to know a molecule that possesses a desirable property and yet not to know how the molecule achieves that effect. The molecule needs improvement (often for things like how well it does its job, or how soluble it is in water) for it to be a realistic drug candidate and we must consider what changes need to be made. In this AI3SD project we know of molecules that efficiently kill the malaria parasite. The molecules need to be improved via small changes to their structures, yet we do not know the molecular biological target of these molecules. Without knowledge of the target it is impossible to design the improvements rationally. In such a project (known as “phenotypic drug discovery”) it is typically the case that the scientist will apply rules of thumb and intuition acquired over many related projects, in order to alter the structure of the molecule in search of those improvements. Yet it is also frequently the case that at such a stage of the project changes can be made to the molecule that accidentally obliterate the desired properties; most typically the molecules lose their potency against the pathogen. We will then have wasted time and resources making inactive molecules. On average each “fail” costs about two thousand pounds to make. It would be much more efficient if we could become more accurately predictive about which molecules need to be made.

In our research consortium, Open Source Malaria (OSM), we have twice tried and failed to generate predictive models. There have occurred, in the period since, major new advances in AI and ML, particularly in the private sector. Since all of OSM’s data and ideas are freely shared in the public domain, it is possible for us to work with anyone in the generation of new models. We have therefore in this project used the AI3SD funds to run a predictive modelling competition and elicited contributions from amateurs and leading AI companies. Several models were better than the others and so, in a crucial part of this project, we asked those winners to predict new molecules, molecules that have never existed before, that they predict will be effective at killing the malaria parasite. We then went to the lab to validate these predictions by making the molecules suggested, and measuring how effective they are at killing the parasite in blood. The result of this was that three of the six predictions were active, a “hit rate” of about the same as the human hit rate across the rest of the project. Interestingly, these actives included a couple of molecules that the human chemists would probably not have tried.

The end result of this work, aside from a new and predictive approach to the synthesis of antimalarial drug candidates, is be a case study of the actual capabilities of new AI/ML technologies in drug discovery: what works, what does not and an examination of why. Notably the project is still ongoing: since all the data and details of the approaches taken are in the public domain, others can try out their own predictive algorithms to see if they can do better.

Text
AI3SD-Project-Series_Report_3_Todd_FinalReport - Version of Record
Available under License Creative Commons Attribution.
Download (305kB)
Text
AI3SD-Project-Series_Report_3_Todd_InterimReport - Other
Available under License Creative Commons Attribution.
Download (218kB)

More information

Published date: 2020
Keywords: AI3SD, Funded Project, Artificial Intelligent, Machine Learning, Drug Discovery

Identifiers

Local EPrints ID: 450088
URI: http://eprints.soton.ac.uk/id/eprint/450088
PURE UUID: 8bcb4b04-167c-47a0-a5f8-8bb9bbd36f59
ORCID for Samantha Kanza: ORCID iD orcid.org/0000-0002-4831-9489
ORCID for Jeremy G. Frey: ORCID iD orcid.org/0000-0003-0842-4302

Catalogue record

Date deposited: 09 Jul 2021 16:33
Last modified: 17 Mar 2024 03:51

Export record

Altmetrics

Contributors

Author: Matthew Todd
Editor: Samantha Kanza ORCID iD
Editor: Jeremy G. Frey ORCID iD

Download statistics

Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.

View more statistics

Atom RSS 1.0 RSS 2.0

Contact ePrints Soton: eprints@soton.ac.uk

ePrints Soton supports OAI 2.0 with a base URL of http://eprints.soton.ac.uk/cgi/oai2

This repository has been built using EPrints software, developed at the University of Southampton, but available to everyone to use.

We use cookies to ensure that we give you the best experience on our website. If you continue without changing your settings, we will assume that you are happy to receive cookies on the University of Southampton website.

×