Witnessing the launch of the Centre for Drug Discovery at the University of Buea, Cameroon
Ersilia has offered training in AI and participated in the foundational workshop of the recently inaugurated Centre for Drug Discovery, supported by the Bill & Melinda Gates Foundation
I believe that every job has its highlight moments of the year, days when the challenges of the preceding months come to fruition and crystallise in an event that is both exceptionally busy and celebratory. We recently experienced one of these moments at the University of Buea, beautifully located at the base of Mount Cameroon, in the South-West region of the country. This region is embroiled in conflict, so the opportunity to engage in academic discussion in this setting is even more significant. The reason for our visit was the establishment of the Centre for Drug Discovery (UB-CeDD), launched with the support of the Bill & Melinda Gates Foundation and under the leadership of Prof. Fidele Ntie-Kang. Fidele is a respected computational chemist, a unique mentor, and a cherished friend. His deep understanding of the research needs in Africa, including practical issues such as delivering reagents, dealing with power outages, and navigating bureaucracy, never ceases to awe me and is a testament to his wisdom, charisma, intuition, and hard work. Fidele’s vision is to create a research institute that will serve as a reference point across the continent. While there is a focus on ethnopharmacology and natural products of plant origin, Fidele also aims for the UB-CeDD to have a global impact, not just a local one. I am humbled by this vision; often, whether intentionally or not, we expect African scientists to focus primarily on local resources and endemic disease areas, setting limitations to their ambitions and scientific freedom. From what I’ve seen, the UB-CeDD closely embodies what a decolonised research agenda should look like. In every context, but especially in this one, Ersilia’s role is to act as mere facilitators, offering expertise and resources to meet the expressed needs of our collaborators, instead of defining the needs ourselves. The new UB-CeDD is running at full capacity, so it was time for us to cement the AI and data science tools that will streamline the production of virtual screening hits in-house. Our team’s perspectives on this can be found in an article we published last year.
A tailored introductory AI course
Our stay began with an introductory course in AI for Drug Discovery (AI4DD). We prepared this eight-day course from scratch to match the specific needs of the UB-CeDD and its emerging regional network. There were twelve participants; some were from Buea, while others came from partner universities from elsewhere in Cameroon, most notably N’Gaoundéré, further north in the country. Later, I learned that one participant had travelled for six days from N’Djamena (Chad) to attend the course. The UB-CeDD had equipped the room with workstations to ensure everything went smoothly. Thus, expectations were high, and the pressure was on us to offer the best possible training. We divided the course in three modules — all materials, including data, code and slides, are available in this repository. We tried to balance introductory notions of Python programming, open-source contribution models, and baseline AI with real-world, state-of-the-art methodologies and advanced AI literature. Our goal was to avoid oversimplification and the endless cycle of basic-level training, which we don’t really believe in, as it is unlikely to yield publishable outputs and is, anyway, already well-covered by fantastic online resources. In other words, the AI4DD course aimed to offer tools that, within a reasonable time, would yield tangible and sound scientific results.
A local implementation of the Ersilia Model Hub
The raison d’être of Ersilia is to offer advanced, yet easy-to-use, AI tools for drug discovery. Therefore, we felt that an effective way to deliver the training was to revolve it around our ever-expanding set of AI tools for antimicrobial drug discovery, collectively referred to as the Ersilia Model Hub. The Ersilia Model Hub can be seen as a marketplace of free, open-source AI models that are ready to deploy without the typical, discouraging challenges related to system setup, code dependencies, etc. We began by showcasing how Ersilia could be helpful for antimalarial drug screening. We selected a few existing models and combined them with auxiliary models related to toxicity prediction, synthetic accessibility of the compounds, and so on. Students were divided into groups and provided with different sets of compounds, some synthetic and some from natural product sources. They were then tasked with selecting, from a list of approximately 500 candidates, a few antimalarial hits that might be intriguing for laboratory testing. In other words, only in a few hours, and from their desktop computers, they were able to shortlist promising antimalarial compounds that could serve as starting-points in a real-world project. We then delved deeper, inviting students to present their current projects related to SARS-CoV-2 (COVID-19), mycetoma and HIV. We scoured the literature and databases to identify relevant training sets for their projects, and collaboratively built three AI models, respectively. These models have since been incorporated into the Ersilia Model Hub and are primed for use in ongoing research at UB-CeDD. The students also had the chance to showcase their first-ever AI models to their peers at the university. Concurrently, the peers had been trained by Dr. Ian Tietjen, from The Wistar Institute in Philadelphia, in the Biomedical Technician Training (BTT) course. So, while Ian provided proficient training in the wet lab, focusing on bioactivity assays, we built data science capacity in the dry lab. We anticipate that both training paths will naturally intersect: experimental scientists will produce data for inclusion in AI models, and, in turn, data scientists will generate predictions to be validated with the pipette. I hope this makes sense.
The inaugural workshop of the UB-CeDD
After the AI4DD and BTT courses, Fidele hosted a workshop in the seminar room of the UB-CeDD. There were about 60 participants (the full room) and a tight program of scientific talks, including presentations by colleagues from Tanzania (Dr. Daniel Madulu Schadrac and Prof. Hulda Shaidi Swai) and South Africa (Prof. Samuel A. Egieyeh), as well as esteemed Cameroonian scientists like Prof. Simon Efange and Dr. Faustin Kamena. Faustin made a case for a truly neglected disease, cryptosporidiosis, for which he has gathered outstanding molecular and microscopy data. Ian presented his convincing work in collaboration with traditional healers in Botswana. Gemma (Ersilia) provided an overview of the existing repertoire of Ersilia projects and their associated challenges. I was in charge of delivering a perspective on AI implemented in resource-limited settings, exemplified by our recent work with the H3D Centre in South Africa. There were chemists in the room, so it was also important to discuss the wonders of generative AI, a field that has kept us busy lately and for which we have already produced some tangible outcomes in the context of Open Source Malaria. Importantly, the stage was offered to junior scientists as well. Prizes were awarded for the best presentations and posters. Overall, the closing ceremony was memorable. It was one of those rare occasions when you are keenly aware that you are witnessing a landmark event for a group of people, and everyone seems to share that sentiment. We can’t wait to return to Cameroon and see where this initial surge of energy has taken Fidele’s team.
