AI Is Speeding Up Drug Discovery — What It Could Mean for Africa

Creating a new medicine is one of the hardest engineering problems humanity attempts.
Scientists first have to understand a disease well enough to identify something inside the body that a medicine could influence. They then search enormous chemical spaces for promising compounds, optimise candidates, test toxicity, move into human clinical trials and eventually convince regulators that a treatment is both safe and effective.

Failure can occur at almost every stage. Artificial intelligence cannot remove that scientific uncertainty. But it is becoming increasingly useful at one crucial task: helping researchers decide where to look before spending years and millions of dollars looking in the wrong place.
That could eventually matter enormously for Africa.

AI Is Compressing the Search Problem

Drug discovery traditionally requires scientists to explore huge numbers of possible biological targets and chemical compounds.
AI can help narrow them. Instead of experimentally testing every possibility, machine-learning systems can analyse biological and chemical data and predict which candidates deserve laboratory investigation first.

One of the most important breakthroughs came from Google DeepMind. In 2020, researchers Demis Hassabis and John Jumper introduced AlphaFold2, which could predict a protein's three-dimensional structure from its amino-acid sequence. By 2024, AlphaFold had predicted structures for virtually all 200 million proteins known to researchers and had been used by more than two million people across 190 countries.

Why does that matter for medicine? Proteins are fundamental machinery inside living organisms. Understanding their structure can help researchers understand diseases and how potential medicines might interact with biological targets.
Work that could previously require years of experimental effort to determine a protein structure can, in some cases, now begin with a computational prediction produced in minutes.
The breakthrough was significant enough that Hassabis and Jumper shared half of the 2024 Nobel Prize in Chemistry, while David Baker received the other half for computational protein design.
That wasn't a Nobel Prize for creating a drug. It was arguably something more foundational: making part of biology dramatically more computable.

An AI-Discovered Drug Has Already Reached Phase 2 Trials

We are also beginning to see what happens when AI moves beyond analysing existing biology into designing actual drug candidates.
Insilico Medicine's rentosertib, developed for idiopathic pulmonary fibrosis, became the subject of a randomized Phase 2a clinical trial. Researchers described both its biological target and small-molecule candidate as having been discovered using generative AI. That is important progress. But we should resist the temptation to declare that AI has already revolutionised medicine.

A drug reaching Phase 2 is not the same as reaching pharmacies. Human biology remains enormously complicated. A molecule can look brilliant on a computer and fail because it is toxic, ineffective in humans, metabolised differently than expected or simply doesn't outperform existing treatment.
The US Food and Drug Administration says it is seeing substantially more drug-development submissions containing AI components across nonclinical, clinical, manufacturing and post-market stages. The regulator had experience with more than 500 submissions containing AI components between 2016 and 2023 alone.

AI isn't replacing clinical science. It is becoming part of it.

Big Pharmaceutical Companies Are Moving In

The latest signal came this month. Novo Nordisk partnered with Anthropic to use Claude in drug research and development, joining pharmaceutical companies exploring AI across scientific workflows.
Anthropic itself has now established a wet laboratory in the San Francisco Bay Area as it expands into biology. Its ambition is particularly interesting: combine AI analysis with physical laboratory experiments rather than treating AI as something that operates separately from scientists and experiments.

Meanwhile, ByteDance recently spun out its AI drug-discovery operation as Anew Labs, which raised $290 million at a reported $1.5 billion valuation. Its work includes biomolecular structure prediction, antibody design and drug discovery.
Companies such as Recursion are taking another approach: generating huge quantities of biological data from experiments and using machine learning to search for patterns that humans would struggle to identify manually.

The direction is becoming clear. Tomorrow's pharmaceutical laboratory may combine human scientists + robotics + biological experiments + enormous datasets + AI models in a continuous discovery loop.

Africa Has More at Stake Than Simply Using the Medicines

Africa's interest shouldn't begin when these drugs arrive at pharmacies. The continent should participate in discovering them.
Tuberculosis provides an obvious example. Around 10.7 million people developed TB globally in 2024, with Africa accounting for approximately 25% of cases. WHO Global Tuberculosis Report

Cancer is another. The International Agency for Research on Cancer estimates that 70% of the world's 81,300 cancer cases attributable to HIV in 2022 occurred in Africa, with cervical cancer and Kaposi sarcoma making up much of that burden.
These are precisely the kinds of problems where Africa cannot afford to remain merely a destination for medicines developed elsewhere.

AI Drug Discovery Is Already Happening in Africa

The encouraging news is that this work isn't confined to California. At the University of Cape Town, the Holistic Drug Discovery and Development Centre (H3D) has built AI and machine-learning tools for drug discovery, including work targeting malaria and tuberculosis. Researchers there developed what they described as the first comprehensive AI/ML virtual-screening cascade deployed in a drug-discovery setting on the African continent.
H3D is particularly important because it isn't simply importing an American chatbot into a laboratory. It combines medicinal chemistry, biological assays, pharmacology and computational drug discovery around diseases including malaria, TB, antimicrobial resistance and HIV.

South African biotechnology company CapeBio is also developing DRUGART, an AI-biological intelligence platform using African microbial datasets to search for potential drugs, biologics, enzymes and diagnostics.
And the nonprofit Magami Open Sciences Initiative is attempting to make AI drug-discovery tools and expertise accessible to African researchers working on diseases that may attract less commercial pharmaceutical investment.

Those efforts matter because diseases affecting poorer populations aren't always the most financially attractive research targets. AI potentially lowers part of the cost of exploring them.

Will AI Actually Make Medicines Cheaper?

This is where optimism needs restraint. If AI reduces years of failed experiments, identifies promising candidates earlier and improves clinical-trial design, pharmaceutical companies can potentially spend less money reaching successful treatments. That could eventually contribute to cheaper drug development.

It does not guarantee cheaper medicine. Patent protection, manufacturing costs, licensing, supply chains, regulatory requirements, market competition, insurance systems and pharmaceutical pricing strategies all influence what a patient eventually pays.

An AI-discovered cancer treatment could therefore be cheaper to develop and still be expensive at the pharmacy. The real breakthrough would come if AI improves success rates as well as speed.
Drug development has historically involved enormous numbers of failures. If researchers can eliminate poor candidates earlier before expensive human trials—the economics of pharmaceutical research could change significantly.

AI's Greatest Medical Contribution May Be Giving Scientists More Chances

The most exciting thing about AI drug discovery isn't that a computer will suddenly invent a cure while scientists watch. It is that scientists may be able to attempt far more possibilities.

• Search more proteins.

• Screen more molecules.

• Understand more genetic differences.

• Investigate diseases neglected because the potential market is considered too small.

• Design treatments around populations historically underrepresented in biomedical datasets.

• And fail earlier, when failure is cheaper.

That last point matters. Every month researchers avoid spending on a molecule that was never going to work is time and money that can be redirected toward another candidate.

For Africa, the opportunity is therefore bigger than waiting for American pharmaceutical companies to produce cheaper medicines. African universities, biotechnology companies and governments should be building the datasets, laboratories, computing infrastructure and scientific talent required to participate directly in AI-assisted medicine.
Because if AI really does compress years of biological searching into weeks or months, the countries that possess both medical data and scientific capability will have an enormous advantage. AI will not make clinical trials unnecessary. It will not make every medicine safe. And it won't automatically make drugs affordable.

But it could allow humanity to explore the enormous search space of biology faster than scientists have ever been able to before.

For patients waiting for treatments that don't yet exist, that extra speed is not merely a productivity improvement. It could eventually mean more lives reached while there is still time to save them.

Read more: Gates Foundation's $1 Billion AI Investment: What It Could Mean for Africa
Read more: Africa Needs Its Own Data Centres But There Is One Major Problem: Power