Course Hones Data Science Skills for Global Health Priorities in Africa

BY: .

A stimulating short course on deep learning with a focus on health data science was hosted by WASHA (Working on Applications for Data Science and Health in Africa) Takwimu (Swahili for Ignite Data), a training grant which is part of the Data Science Initiative (DSI)-Africa consortium.

The consortium is made up of institutions across the African continent with a focus on harnessing data science for health discovery and innovation in Africa. This grant is based at UKZN’s Centre for Rural Health (CRH) in the School of Nursing and Public Health (SNPS) with collaboration from the Schools of Mathematics, Statistics and Computer Science; and Agricultural, Earth and Environmental Sciences.

Funded by the National Institutes of Health (NIH U2RTW012140), the award is led by the Harvard T.H. Chan School of Public Health in collaboration with the Heidelberg Institute of Global Health (HIGH). UKZN is the programme’s hub in Africa that focuses on research training to harness health data science for global health priorities: health systems strengthening, and food systems, climate change and planetary health.

This training was attended by delegates from the four spoke partners in sub-Saharan Africa: Nigeria (University of Ibadan), Uganda (Makerere University), Tanzania (Muhimbili University), and Ghana (University of Ghana) as well as a number of South African delegates. The grant aims to build on existing data science research capacity at the partnering African institutions to enhance innovative new health data science research capacity.

UKZN Co-Principal Investigator, Professor, Henry Mwambi of the School of Mathematics, Statistics, and Computer Science explained that biomedical research generates a lot of data, which is often varied. ‘Because of the data’s complexity, data science methods help us to understand what it is indicating within a particular domain, in this case, about a particular health problem,’ he said. ‘Huge volumes of data cannot be analysed by a human or by virtue of a simple algorithm, and this is where artificial intelligence (AI) and machine learning methods come in. Deep learning is an extension to reinforce existing machine learning methods to enhance their predictive capacity.’

Mwambi and senior lecturer and academic leader for Computer Science, Dr Mandlenkosi Gwetu, who co-facilitated a week-long virtual bootcamp prior to the in-person training, cited the example of pandemics, that call for appropriate predictive learning models. ‘Computational methods are needed to process data such as genomic sequencing data which generates tens of thousands of genes that are predictive of a certain disease or diseases. It is important to boost already existing methods,’ said Mwambi.

Deep learning, a subfield of machine learning that builds predictive models using large artificial neural networks, has revolutionised the fields of computer vision, automatic speech recognition, natural language processing, and numerous other areas including public health, medicine and computational biology.

The bootcamp introduced participants to Epidemiology 101 and Python - a popular general-purpose interpreted, interactive, object-oriented, and high-level programming language. The epidemiology part of the boot camp was led by Professor Palwasha Khan of the London School of Hygiene and Tropical Medicine who is also a researcher at the Africa Health Research Institute (AHRI) assisted by Dr Stephen Olivier of the AHRI. It focused on the basic concepts of epidemiology in global health, deep neural networks, basic neural networks, convolutional neural networks and recurrent neural networks’ structures. It also examined biomedical and public health applications. Participants were expected to be familiar with calculus, linear algebra, machine learning and Python.

Harvard T.H. Chan School of Public Health’s Lead Facilitator, Professor Santiago Romero-Brufau introduced deep learning, differentiating the methodology from machine learning, the subset of AI that focuses on building systems that learn - or improve performance - based on the data they consume. The practical sessions were led by Dr Mohanad Mohammed, a postdoctoral fellow at UKZN funded by the DSI-Africa consortium funding from the NIH.

A number of local and international experts presented at both the bootcamp and the short course, with participants reporting that this made for a vigorous and insightful learning experience.

UKZN DSI-A Training Director, Professor Saloshni Naidoo said, ‘Building data science capacity in South Africa and Africa is an important initiative. The introductory trainings were the scaffolding needed for this module that leads well into the domain-specific areas of health systems strengthening and climate change and health. Participants acquired new and much needed-skills which can be absorbed into our research agendas.’

Words and photographs: Lunga Memela