Scientists looking for the next big thing in malaria treatments have plenty of avenues to explore, with thousands of potential anti-malaria chemical compounds available for study.
Jessica Thibaud, a doctoral candidate at Stellenbosch University in South Africa, is among the growing body of researchers turning to AI and machine learning. She is developing a computational model that would help researchers filter large amounts of possible compounds and identify those most likely to suit their specific approach to tackling malaria.
For a prototype model, Thibaud is exploring compounds able to short-circuit an enzyme known as Plasmodium falciparum cGMP-Dependent Protein Kinase (PfPKG). PfPKG plays an important regulating function in the life cycle of the malaria-inducing parasite species known as Plasmodium falciparum. This is the deadliest of the four Plasmodium species known to cause malaria in humans via the bite of the female Anopheles mosquito. The species accounts for the vast majority of malaria cases and deaths in Sub-Saharan Africa. Inhibit the function of this enzyme, the logic goes, and the development of the parasite is thwarted.
“It’s an interesting target as it’s resistant to resistance,” explains Thibaud, citing a 2020 study that found cGMP-dependent protein kinase (PKG) to be ‘resistance-refractory’.
The researchers concluded that although the parasite is able to gain resistance to drugs that inhibit PKG through other means, the PKG enzyme itself never mutated. “Which is what makes it such a promising target for drug discovery,” she says.In developing her computational model, Thibaud started by downloading a handful of ‘virtual’ libraries and investigating the compounds in each using a machine-learning technique known as principal component analysis (PCA). PCA is able to filter through large datasets, reducing large amounts of variables to more manageable dimensions, while not losing too much of the dataset’s original richness or variance.
This allowed her and then collaborator Dirkie Myburgh, at the time a PhD student at Stellenbosch, to develop a 2D map of anti-plasmodium ‘chemical space’, essentially a cloud of promising molecules. The most promising ones are then visually highlighted.
“What we predict is that the probability of identifying an inhibitor of our target – PfPKG – is much higher when we select compounds from the region that we identified as being a region of enrichment, rather than going somewhere completely random on the map,” explains Thibaud.
However, that process reduced the initial libraries of some 50,000 compounds to just over 6,000. “But we can’t buy 6,000 compounds,” says Thibaud. For perspective, the recent purchase of 10 compounds, with around 5mg of each, cost around $US 1,300).
Dock and block
To bring 6,000 down to a more manageable number, Thibaud was specifically on the lookout for molecules that could dock and block the enzyme at the ATP-binding site. ATP is critical to the storing and transferring of energy in cells. “So if a different molecule comes along and blocks off that site, the enzyme can no longer perform its functions and the parasite life cycle is interrupted,” Thibaud explains.
Thibaud was able to whittle the 6,000 down to around 500 before going through each manually.
That exercise revealed 41 compounds, including inactive compounds that would serve as negative controls for validating her model. Further enzyme testing is being conducted by the Holistic Drug Discovery and Development Centre (H3D) at the University of Cape Town.
This in turn, revealed three compounds with good activity against the PfPKG enzyme. (Until publication, these compounds cannot be named). At least two are current FDA-approved anti-cancer drugs.
“I’m very interested in drug repurposing, where you look at existing medicines that treat other diseases, but can also be applied to malaria,” says Thibaud.
This is essentially a cost-saving approach, she explains. “If it’s an existing drug, then all the toxicity, solubility and dosage studies have already been conducted.”
Repurposing for malaria, for instance, may ‘simply’ require a change in the dosage regime. These are avenues that will now have to be explored in further research.
Exploring other compounds
Thibaud is hoping her machine learning model can help identify compounds that could hit multiple targets in the parasite, in this way disrupting multiple disease pathways, an approach known as polypharmacology.
Literature suggests that certain compound scaffolds could impact both the workings of PfPKG and the formation of hemozoin, explains Thibaud. Hemozoin is a waste product formed via what’s called the ‘heme detoxification pathway’ during the parasite’s blood cycle, i.e. the growth of successive broods of parasites within red blood cells. The production of hemozoin is essential for malaria parasites as it protects them from a build-up of molecules known as free heme, which are toxic to the parasites.