With the rapid emergence of resistance towards current antimalarial drugs, the progress made in eliminating malaria has come under threat. This highlights the need for new antimalarial drugs with novel modes of action. However, the process of antimalarial drug discovery is time consuming and resource intensive and relies on resolving complex biological problems associated with drug and drug target identification.

Machine learning (ML) has fast become an important tool to aid and enhance drug discovery, due to the ability of these computational approaches to rapidly and accurately identify trends within large and inherently complex datasets. ML has already played a major role in cancer drug discovery, by aiding in decision making to reduce failure rates in drug discovery. Two recent papers from postgraduate students in the Malaria Parasite Molecular Laboratory (M2PL) in the Department of Biochemistry, Genetics and Microbiology under SARChI Chair Prof L Birkholtz, has made major breakthroughs in the use of ML in antimalarial drug discovery. The M2PL forms part of the Parasite Control Cluster of the University of Pretoria Institute for Sustainable Malaria Control (UP ISMC).

Diagram depicting Machine Learning to aid and enhance drug discovery.

PhD student, Rudi van Wyk, published a novel web-based tool that used ML to better understand the parasite’s biology and its’ gene regulatory mechanisms, in the hopes of identifying promising drug targets (van Wyk et al., Malaria Journal, (2021) 20:317). Following a series of publications focused on resolving gene regulation and subsequent drivers of the parasite development, Rudi constructed a tool to spur research into gene regulation even further titled: “MALBoost: a web-based application for gene regulatory network analysis in Plasmodium falciparum”. The parasite’s complexity is perhaps best described by its great many “unknowns”, with multiple different development stages spanning across two host species, with the many unknown elements driving differentiation and proliferation remaining a mystery. Deconvoluting these mysteries to gain an understanding of the parasite biology and ultimately greater insights into the effect of antimalarials at a molecular level (mode of action), requires an equally complex tool. MALBoost therefore perfectly merged ML power to resolve biological complexity and promises to be a very useful tool to the malaria community in search of novel drug targets. 

PhD student, Ashleigh van Heerden, used a different approach and published the discovery of biomarkers and chemo-transcriptomic fingerprints to indicate the mode of action of antimalarial drug candidates. Her paper: “Machine learning uses chemo-transcriptomic profiles to stratify antimalarial compounds with similar mode of action” was published in a special issue of Frontiers in Cellular and Infection Microbiology (van Heerden et al., Front. Cell. Infect. Microbiol. 11:688256). The power of ML was used in this study to interrogate patterns in drug-induced expression profiles of the parasite. Ashleigh was able to generate robust ML models that could stratify compounds based on gene expression profiles, for a minimum set of informative biomarkers, a novel approach in antimalarial drug characterisation. The applications of these findings are vast and will allow the rapid characterisation of the mode of action of antimalarial drug candidates, and as such, advance the initial stages of a drug discovery programme.

The expertise that is being built within malaria research and using cutting-edge developments like ML indeed strengthens our position as leading innovations and efforts in the fight against the disease in South Africa.