From code to care: Dr Musula Sinkala on the role of machine learning in healthcare
When Musalula Sinkala graduated with a degree in molecular pathology and biochemistry, he hoped to conduct research that would improve human well-being. However, as a young graduate, he found himself bogged down by repetitive, time-consuming lab experiments.
“We were looking for proteins that could be used to diagnose a particular disease,” recalls Sinkala. “And the specific proteins were a few among hundreds – so we had to conduct hundreds of experiments. I suggested to my professor that maybe there’s a way of narrowing down the research space, such that you only have a few proteins we could try out in the experiments.”
After doing some background reading, Sinkala discovered researchers using systems biology methods in the way he had imagined: modelling potential outcomes to eliminate non-viable options without trial and error. Inspired, Sinkala embarked on a journey to learn programming. Ten years later, with a doctorate in bioinformatics, Sinkala is now a lecturer and research fellow at the University of Cape Town. Here, he combines his biology and machine learning knowledge to accelerate healthcare research and inspire a new generation of African health researchers.
“What drives me, I think, is curiosity and understanding things,” he says. “In essence, it’s a search for truth or for facts that can lead to the betterment of people’s lives.”
The GIFT device
As a key member of the GIFT team, Sinkala looks forward to contributing his problem-solving skills to the challenge of screening women for genital inflammation, an indication of bacterial vaginosis (BV) and sexually transmitted infections (STIs). Despite their ill effects on women’s health, many cases of BV and bacterial STIs lack apparent symptoms and are left untreated.
Sinkala and the GIFT team hope to change this by using artificial intelligence to develop a screening device for wide distribution in low-income settings. Now being finalised, the GIFT device tests for three biomarkers of vaginal inflammation. “In essence, my role is to identify cutoff points for those [biomarkers] that would allow us to identify women who will be positive for inflammation,” Sinkala explains.
Because of the combinations of biomarkers, identifying the best levels would be time-consuming if manually calculated. By applying statistical methods and machine learning to data collected from the three clinical sites, Sinkala’s involvement with ensure a speedier finalisation and roll-out of the device. H
His expertise demonstrates how machine learning can help bridge the gap between complex biological data and practical healthcare solutions. The GIFT project has meant “moving from sitting alone in an office writing code all night to working on something that has an immediate impact on the community and women.”
Machine learning in healthcare
Sinkala’s work with the GIFT team illustrates how machine learning aids healthcare research. “The applications of machine learning in healthcare today are almost limitless,” he says.
Machine learning has become essential for making sense of the vast datasets generated in biological research. “It would take us months or years to sift through them [datasets] and gain insights. But you can train machine learning models now from those data sets, and those models can give us insight into information that could be actionable.”
“My North Star has been trying to use machine learning to improve healthcare by minimising this space of things we’re looking at,” says Sinkala. “For the GIFT device, it’s to accelerate the indication of cytokine cut-off points. And for other areas where I’m involved, for example, in cancer research, it’s accelerating the discovery of targets presented as this type of cell or that cell that could be used to develop specific treatments.”
Future applications of machine learning
Sinkala believes machine learning can also drive a more proactive, preventative approach to healthcare. With digitised healthcare records, machine learning models can help to identify patients at risk for a particular disease, so that healthcare professionals can intervene before they develop.
“If you’ve got a database of patients and their blood profiles and other clinical characteristics, you can begin to train models on those datasets or clinical features to predict who is more likely to develop heart disease, and you will have a probability attached to that.”
Sinkala finds the predictive power of machine learning its most exciting aspect. “When you make accurate predictions, you get fewer people reporting at the hospital, fewer occupied beds, and lower expenses,” he explains. By combining predictive power with screening tools like GIFT, Sinkala hopes that healthcare can shift towards timely interventions that significantly improve patient outcomes.
For Sinkala, sharing machine-learning knowledge through his role in the GIFT project is as important as applying it. “If more people get to know how to do this work, they don’t even need me to be there,’ he explains. “For me, that’s very important – not just me doing it, but other people learning how to do it.”
Sinkala’s advice for young researchers is to trust their instincts and challenge norms. When he first suggested using machine learning to accelerate health research as a young graduate, his professor dismissed the idea as crazy. “When people tell you something can’t be done, but you feel it could be, try it out… Become curious, and don’t always trust the expert.”
His journey showcases how bold ideas and new technologies can transform healthcare. By supporting tools like the GIFT device and encouraging the next generation of curious thinkers, we can realise a future where timely, accessible interventions save countless lives.