AI-Driven Drug Discovery for Neglected Tropical Parasites

Background

Leishmania parasites (and the related Trypanosoma species), are together the deadliest single-cell parasites after the malaria parasite. The World Health Organisation recognizes them as key neglected tropical diseases – we want to address this. 

Target-based drug development has challenges because these parasites are evolutionarily divergent from humans and model organisms. Around half of the genome encodes ‘hypothetical’ proteins; parasitespecific proteins with no ortholog in model organisms which allows prediction of function. There may be many undiscovered good, highly specific, drug targets in this set. On host side, these parasites affect large tracts of the world covering huge human genetic diversity and computational drug screening must account for risk of off target effects in this diversity of people. 

We have generated foundational data for this project: the first genome-wide analysis of proteins vital for the Leishmania to infect mammals [1]. In combination with improved quality of AI protein structure predictions for these parasites [2], this is the ideal time to tackle the drug discovery pipeline which could then be generalised to other unicellular parasites: Plasmodium, Toxoplasma, Giardia, etc. 

Aims 

The project wil develop an end-to-end digital drug target and small molecule discovery pipeline, incorporating cutting-edge molecular biology, AI, and bioinformatic approaches, including protein structure prediction, drug binding pocket identification and small molecule docking targeting parasite proteins. To avoid off-target effects, off-target binding pockets in human proteins will be similarly be analysed, incorporating human polymorphisms common in affected areas. Programmatic use of local large language models (LLMs) will be used for automated data interpretation. A first proof-of-concept has been implemented, but its full potential is far from being maximised. The project will lead to predictions of 1) novel drug targets and 2) candidate small molecules which are predicted to bind to these targets. The project is planned to be primarily computational, with a recommended experimental phase testing identified commercially-available compounds in parasite toxicity experimentally. 

This will build from existing phenotyping data, analysing 5 replicates of 10 different experimental assays for all ~9,000 genes in the parasite genome. And, will be further supported by our current [1] and previous [3] high-content image analysis of parasite subcellular localisation. 

Specific objectives: 

1. Develop a robust pipeline to identify novel drug targets from phenotypic data, predicted protein structure and binding pocket potential. 
2. Implement a pipeline for in silico screening of small molecules targeting these novel targets. 
3. Experimentally validate commercially available lead compounds discovered.

Training outcomes 

The student will gain extensive training in applied application of local and remote advanced AI tools, protein and protein structure bioinformatics, data science/statistics and cell biological data science. The student will be embedded in a research group and research environment with extensive experimental parasite work, while also being directly supervised by a PI experienced in developing such computational pipelines. This will train the student to use real noisy biological datasets in key biomedical informatics areas, ideal for future work in AI drug discovery of key interest to academic and industry research. 

References 

[1] LeishGEM Team (Sidonie Aellig, Karen Billington, Jeziel D. Damasceno, Laura Davidson, Ulrich Dobramysl, Ruth Etzensperger, Eden Ramalho Ferreira, Eva Gluenz, Jeremy C. Mottram, Rachel Neish, Raquel Pereira, James Smith, Jack D. Sunter, Petr Volf, Richard J. Wheeler, Matthew Young) (2024) “LeishGEM: genome-wide deletion mutant fitness and protein localisations in Leishmania.” Trends Parasitol 40(8):675-678 doi:10.1016/j.pt.2024.06.003 https://www.leishgem.org/ 

[2] Wheeler RJ (2021) “A resource for improved predictions of Trypanosoma and Leishmania protein three-dimensional structure.” PLoS One 16(11):e0259871 doi:10.1371/journal.pone.0259871 https://www.ebi.ac.uk/about/news/updates-from-data-resources/alphafold-database-communitydatasets/ 

[3] Billington K, Halliday C, Madden R, Dyer P, Barker AR, Moreira-Leite FF, Carrington M, Vaughan S, Hertz-Fowler C, Dean S, Sunter JD, Wheeler RJ, Gull K (2023) “Genome-wide subcellular protein map for the flagellate parasite Trypanosoma brucei.” Nat Microbiol 8(3):533-547 doi:10.1038/s41564-02201295-6 https://www.tryptag.org/ 

[4] Center for AI Safety, Scale AI, HLE Contributors Consortium (2026) “A benchmark of expert-level academic questions to assess AI capabilities.” Nature 649(8099):1139-1146 doi:10.1038/s41586-02509962-4