Systemic bias and poor profit margins are key reasons why infectious diseases, such as malaria and tuberculosis, remain the top killers in the global south.
These diseases account for a huge six of the top 10 causes of death in developing countries, but a mere ten percent of the drugs currently in development are intended to target them, according to the World Health Organization.
In other words, not only do people in these regions experience systemic barriers to accessing equitable healthcare, but scientific and biomedical research is subject to them too. A key problem is that the vast majority of such research activity is concentrated in high- and upper middle-income countries. In fact, lower income countries publish only five percent of the world’s scientific papers.
As a result, the focus is inevitably on tackling the global north’s most pressing regional health concerns, such as heart disease, rather than on eradicating the communicable diseases posing the biggest challenges in the global south. This situation is reinforced by the fact that the pharmaceutical industry tends not to consider the return on investment from drug discovery in this field to be significant enough.
But it also begs the question of how best to empower local researchers to develop their own solutions to local problems, for instance, by making the most of natural resources and local expertise in traditional medicine and plant-based treatments. One organization that is trying to answer that question is Ersilia.
The transformational nature of data science
The charity was set up as in the UK in November 2020 by three early-career scientists: Gemma Turon, a molecular biologist and now Chief Executive; Miquel Duran-Frigola, a computational biologist, Chief Scientific Officer and Trustee, and Edoardo Gaude, a Trustee who also has a PhD in oncology.
After volunteering and sharing his data science expertise with local hospitals in Mozambique andZambia during sabbaticals there, Duran-Frigola had realized that:
Data science in those contexts where funding is very limited can be transformational. So, when I came back to Europe with Gemma, we started Ersilia with Edoardo, who was based in Cambridge, to provide data science support to research centres and hospitals in the global south, mainly sub-Saharan Africa. Rather than working with clinics, we decided to focus on drug discovery, a specific area of global health that’s closer to the research side we’re familiar with.
The aim here was to create ready-to-use Machine Learning (ML) models to help speed up experimentation and cut the development costs for drugs that fight against infectious and neglected tropical diseases, such as Chagas. This work was, and is, based on three key pillars:
Open source and the open science model
Because the people the charity works with have limited funds, its code is open source and comes free-of-charge to remove any barriers to access. There are no patents either, which Duran-Frigola believes is “very important to further global health research”.
Researchers in institutions in the global south have an in-depth understanding of local problems. So, rather than simply impose technology on them, the focus is on working with them in situ to jointly implement solutions.
As Ersilia consists of only a small in-house team of two (plus hundreds of volunteers), the aim is to ensure the work can continue in situ once they leave. As a result, knowledge transfer and training in how to use data science tools is considered key. The machine learning models also require low resources and can be used without an internet connection.
To store its ML models centrally, meanwhile, the organization created a Linux-based hub to act as a repository. The first phase of development involved populating it with previously validated models to build up trust. The second step continued this approach by identifying sound publicly available data sets on specific diseases, so the charity could build its own models in-house.
As a result, any scientist or clinician can now browse the hub and use any one of the 110 models stored there to undertake predictive modelling without needing to either write their own code or pay for software licences. But there are other benefits too, as Turon explains:
If you’re developing a new drug for disease, you test X number of molecules before moving onto computer models, animal models and then clinical trials. If the drug isn’t good enough at any stage, it won’t progress, but if it fails at the clinical phase it’s very costly as it’s taken years to get there. So, what AI does is give us predictive power. For example, it predicts this will be more effective against pathogens and that will interact with other drugs. So, with AI you can select molecules that have the highest chance of success and focus your efforts there.
Practicing global collaboration
The third phase of development though involves collaborating with research institutions in developing world countries to build specific tailored models. One such institution is the Open Source Malaria consortium.
It has used a series of newly-developed models to predict new drug candidates and tested eight of them in vitro in its labs. Some 50% of the molecules under investigation came out as active against malaria compared with the usual 15% or less, boosting the potential success rate significantly.
Another beneficiary is Africa’s leading drug discovery centre, the H3D Centre in Cape Town, South Africa, which employs 100 or so scientists. Fifteen models based on 10 years-worth of data about 10,000 molecules have just been implemented here, with the aim of boosting efficiency.
A third customer is the University of Buea in Cameroon, where Ersilia is helping to kickstart its new data science activities. Discussions on suitable models will start in September and the plan is to roll the technology out over the next five years.
Another goal is to replicate this collaborative approach at another dozen centers across sub-Saharan Africa and Latin America within three years, becoming a “reference center” for such activity in the process, Duran-Frigola says.
Ensuring a sustainable future
As for where the charity derives its funding to ensure its activities are sustainable into the future, this comes from a variety of sources. These include research grants from organizations, such as the Welcome Trust and European Union, crowdfunding from the open source community, and corporate donors, which include tech companies.
One such company is data management software provider, Splunk, which selected the charity to participate in its Global Impact scheme. Ersilia had in the past analyzed and managed the data relating to its ML models manually, which was a costly and time-consuming process.
But by rolling out Splunk’s Enterprise system at the end of last year, this process has now been automated, making it possible to monitor and control it more effectively. For instance, the system’s dashboard provides statistics on model usage and the overall health of the hub, sending alerts if things go wrong.
It also offers data on the activities of the open source developer community. About 120 members regularly check that the models, and the automated workflows they create, work effectively once they have been published. They likewise upgrade and maintain the hub and related infrastructure. The organization also has about 100 GitHub contributors.
By introducing the Enterprise system, the charity has been able to save 700 hours of manual work in 10 months and quintuple the number of ML models it has created this year to 100. By the end of next year, the aim is to boost this figure to 500. As Duran-Frigola concludes:
Infectious disease is a global health problem, especially with the growing emergency of anti-microbial resistance. So, I can’t see our focus moving beyond infectious and neglected tropical diseases in the next 10 years. As for any pushback from big pharma, it’s more about cooperation than rivalry. They play an important role in global health and we’re trying to cover some gaps in areas that haven’t been covered enough, so it works well.
Ersilia is definitely one to watch in a global health context. Its collaborative open science/open source approach to drug discovery for developing countries with limited resources is inspired and, all things being equal, could make a lasting mark on the lives of millions of people.