At a glance
The Ersilia Open Source Initiative, a Catalan nonprofit, runs the Ersilia Model Hub, a curated collection of about 150 free AI and machine learning models that predict how candidate molecules might treat infectious and neglected diseases. Researchers in the Global South can find, run and contribute models without data science teams or expensive lab experiments, closing a resource gap that leaves diseases of poverty understudied.
- Developing a single new drug takes on average 10 years and 1.3 billion dollars, so pharma prioritizes diseases of wealthy countries
- Only 15 percent of drugs in development worldwide target infectious diseases, according to the WHO Global Health Observatory
- The Ersilia Model Hub offers about 150 pre trained models that run free through a simple interface or a command line tool
- Ersilia has directly trained more than 100 researchers through in person workshops in South Africa, Zambia, Cameroon and Ghana
- Its biggest partner is the H3D Centre at the University of Cape Town, Africa's first integrated drug discovery and development center
The Ersilia Open Source Initiative is a Catalan nonprofit that gives away the tools of modern pharmaceutical research to the scientists who need them most. Its central project, the Ersilia Model Hub, is a curated collection of about 150 artificial intelligence and machine learning models that predict how candidate drug molecules might behave against infectious and neglected diseases. Anyone, anywhere, can use the hub for free. That matters because the diseases it targets, malaria, tuberculosis, leishmaniasis, Chagas disease and dozens of others, largely affect people whose markets pharmaceutical companies have decided are not worth the investment.
The numbers behind that imbalance are stark. Bringing a single new medication to market takes an average of 10 years and 1.3 billion dollars, according to a study published in Nature Communications. With stakes that high, companies chase returns in wealthy countries. The World Health Organization's Global Health Observatory estimates that only 15 percent of drugs in development worldwide target infectious diseases at all, a category that includes six of the top ten causes of death in low income countries.
Why Drug Discovery Needs AI, and Why Access Is Unequal
Machine learning has transformed early stage drug discovery. Instead of synthesizing and testing every candidate molecule in a lab, researchers can run computational simulations that predict whether a compound is likely to inhibit a pathogen, cross cellular membranes, or turn toxic. These models are cheap to run compared with experiments, which makes them ideal starting points for labs with limited budgets.
The catch is that using them normally requires data science expertise, expensive computing infrastructure, or both. A chemist in Cameroon studying malaria parasites often has neither. Meanwhile, research produced in the Global South amounts to less than 5 percent of the world's scientific output, a gap driven by exactly these missing resources. Ersilia's founders, three biomedical researchers named Miquel Duran-Frigola, Gemma Turon and Edo Gaude, built the organization to close that specific gap.
A Model Hub Built Like Open Source Software
The Ersilia Model Hub works like an app store for drug discovery models. Researchers can browse a catalog, pick a model, and run predictions on their own candidate molecules through a browser interface that requires no coding, or through a command line tool for those who want more control. The models come from published academic research and from Ersilia's own scientists, all packaged and tested so they actually run.
Ersilia went further and automated the packaging pipeline using GitHub Actions, working with GitHub engineers to build a self service process where contributors submit models and the system validates and publishes them. A model that once would have lived buried in a paper's supplementary code now runs on public infrastructure that any researcher can access for free. The approach earned Ersilia recognition in Fast Forward's AI for Humanity Report 2026, which profiled it among nonprofits doing substantive work with AI rather than pilots built for press releases.
Training People, Not Just Shipping Software
Software alone does not create research capacity, and Ersilia treats training as a first class part of its mission. The organization has directly trained more than 100 researchers, including chemists, biologists, pharmacists and bioinformaticians, through in person workshops on applying AI to drug discovery held in South Africa, Zambia, Cameroon and Ghana. Its strategic plan for 2025 through 2027 puts capacity building at the center, arguing that lasting impact means partner institutions grow their own programs rather than depending on outside experts.
The hub's biggest partner is the Holistic Drug Discovery and Development Centre at the University of Cape Town, known as H3D. Founded in 2010 as Africa's first integrated drug discovery and development center, H3D uses Ersilia's tools to advance homegrown drug discovery on the continent. Ersilia has also set a goal of establishing the hub in 15 institutions across Africa, an effort to embed the technology where the diseases actually are.
Open Science as a Structural Answer
What makes Ersilia unusual is not just the technology but the posture. Everything is open source and open access. Models, code, documentation and training materials are public goods by design. The organization's stated principles frame this as a response to the extractive patterns that have historically defined global health research, where samples and data flow from the Global South to institutions elsewhere and the benefits rarely flow back.
The model is also unusually lean. A nonprofit with a small team can maintain a hub of 150 research grade models because it stands on open infrastructure and a volunteer friendly contribution process. For funders looking at the space, Ersilia represents the unglamorous middle of AI for good, not a splashy demo but plumbing that makes existing science usable by the people closest to neglected diseases.
A researcher in Lusaka or Yaounde testing whether an existing compound might work against a local pathogen no longer needs a supercomputer or a contract with a software vendor. They need a laptop, an internet connection, and the Ersilia Model Hub. That is what redistributing the tools of discovery looks like in practice.
Common Questions
What is the Ersilia Model Hub?
The Ersilia Model Hub is a curated, open source collection of about 150 pre trained AI and machine learning models for infectious and neglected disease drug discovery. Researchers can run models through a browser interface with no coding required, or through a command line tool, all for free.
Who founded Ersilia?
Ersilia was founded by three biomedical researchers: Miquel Duran-Frigola, Gemma Turon and Edo Gaude. It is registered as a Catalan nonprofit organization and works with drug discovery centers across the Global South.
Why do pharmaceutical companies neglect tropical diseases?
Developing a new drug takes on average 10 years and 1.3 billion dollars, so companies focus on diseases with profitable markets in wealthy countries. Only 15 percent of drugs in development worldwide target infectious diseases, even though they include six of the top ten causes of death in low income countries.
How does Ersilia support researchers in Africa?
Ersilia has trained more than 100 researchers through in person AI and drug discovery workshops in South Africa, Zambia, Cameroon and Ghana, and aims to establish the Model Hub in 15 institutions across Africa. Its largest partner is the H3D Centre at the University of Cape Town, Africa's first integrated drug discovery and development center.
How do the AI models help drug discovery?
The models run computational simulations that predict how a candidate molecule might treat a particular disease, whether it will inhibit a pathogen or how it will behave in the body. These predictions are far cheaper than lab experiments and give researchers in low resourced settings a way to prioritize which compounds to synthesize and test.
Sources: Ersilia Open Source Initiative strategic plan 2025 to 2027; The GitHub Blog, Automating open source: How Ersilia distributes AI models to advance global health equity (August 2024); Nature Communications, investments and timeline for drug development; WHO Global Health Observatory via Ersilia strategic plan; Fast Forward AI for Humanity Report 2026; University of Cape Town H3D Centre.
