Fifty years. Thousands of scientists. Billions in research funding. The greatest unsolved problem in biology — the protein folding problem — resisted every approach, every model, every mind that was brought to bear on it.
Then, in 2020, an AI called AlphaFold solved it. Not approximately. Not partially. Completely.
Medicine has not been the same since.
The Problem That Stumped Everyone
Think of a protein as a string of differently shaped beads — amino acids — that spontaneously twist and fold into a precise three-dimensional structure the moment they are assembled. That final shape is determined entirely by the sequence of amino acids. But predicting what shape any given sequence would fold into was, for half a century, beyond human ability.
Why does the shape matter? Because a protein's shape determines its function. Haemoglobin carries oxygen because of a specific shape. Insulin tells your cells to absorb glucose because of a specific shape. When a protein folds incorrectly — misfolding — the result can be catastrophic: Alzheimer's disease, Parkinson's, cystic fibrosis.
The international competition to solve protein folding was called CASP — Critical Assessment of Structure Prediction. Every two years, research teams submitted their best predictions. Progress was real but slow. Then in 2020, AlphaFold entered CASP14 and didn't just win — it made every other submission look like it was from a different era. (Source: CASP14 results, November 2020)
What AlphaFold Did
DeepMind's AlphaFold used a novel neural network architecture that combined evolutionary information with attention mechanisms that learned which amino acids in a sequence tend to be near each other in the final folded structure. The results were extraordinary: prediction accuracy comparable to expensive experimental methods that take months in a laboratory.
DeepMind then released the AlphaFold database publicly: over 200 million protein structure predictions, covering essentially every protein of every known organism on earth. Free. Accessible to any researcher in the world. (Source: DeepMind AlphaFold database announcement, July 2022)
The scientific community reacted with something close to disbelief. Work that would have taken individual laboratories decades was now available to download. New protein structures that would have required months of crystallography experiments could be modelled in hours.
"A problem that occupied the greatest minds in biology for fifty years became a solved problem on a Tuesday in November. Science has not had a moment like this in a generation."
What This Means for Drug Discovery
Developing a drug is fundamentally a problem of finding a molecule that fits a specific protein — like finding a key for a lock. For fifty years, drug developers were trying to make keys without being able to clearly see the locks. AlphaFold changes this.
When you can see the protein structure in precise detail, you can computationally design candidate drug molecules that fit the target protein, screen millions of potential candidates virtually before any laboratory work begins, and identify binding sites that were previously unknown.
The timeline from target identification to clinical candidate is compressing. Several drug candidates identified using AlphaFold structures have already entered clinical trials. (Source: DeepMind / Science press reporting on AlphaFold clinical applications, 2023-2024)
Rare Diseases and the Forgotten Patients
A rare disease affects fewer than 1 in 2,000 people. The commercial incentive to develop treatments is limited — the patient population is too small to justify the cost of traditional drug development.
The result is that of the approximately 7,000 known rare diseases, only about 500 have approved treatments — a figure cited by the National Center for Advancing Translational Sciences (NCATS) and the rare disease advocacy organisation Global Genes. The rest are diseases where patients receive a diagnosis and then are largely left to manage with what exists.
AlphaFold changes the economics. When the expensive, time-consuming early research phase of drug discovery becomes dramatically faster and cheaper, the calculus for rare diseases shifts. Academic researchers can now pursue rare disease targets that would have been impractical before.
The Antibiotic Resistance Crisis
Antibiotic resistance is one of the most serious public health threats of our time — classified as such by the WHO. Bacteria evolve resistance to existing antibiotics faster than new ones are developed, and the pipeline of new antibiotics has been nearly empty for decades.
AlphaFold is helping address this in two ways. First, by revealing the structures of proteins essential to bacterial survival with no human equivalent — ideal drug targets. Second, by enabling the mapping of resistance mechanisms at the protein level, revealing how to design drugs that are harder to become resistant to.
Researchers have already used AlphaFold to identify novel antibiotic candidates targeting bacterial proteins that were previously invisible to drug design. (Source: Science, reporting on AlphaFold antibiotic research applications, 2023)
AI as Lab Partner, Not Just Tool
AlphaFold is the most visible example of a broader shift happening in scientific research. AI is not just processing data faster — it is generating hypotheses, suggesting experiments, and accelerating the entire scientific method.
AlphaFold 3, released in May 2024, extended the capability beyond proteins to predict the structure of DNA, RNA, and how they interact with proteins and with small molecules — which is exactly what drugs are. (Source: DeepMind AlphaFold 3 announcement, May 2024, published in Nature)
Beyond AlphaFold, AI systems are being used to design entirely new proteins that do not exist in nature, optimised for specific functions — proteins designed to break down plastics, deliver drugs to specific tissues, or act as more effective enzymes in industrial processes. Biology is becoming programmable in a way that was previously theoretical.
What This Means For India
India carries a disproportionate burden of several diseases particularly amenable to AlphaFold-accelerated drug discovery. Tropical diseases — kala-azar, dengue, malaria — that historically received insufficient pharmaceutical investment because they primarily affect low-income populations are now being targeted by researchers using AlphaFold structures.
India's pharmaceutical industry is the world's largest generic drug supplier. The opportunity to use AI-accelerated drug discovery to develop novel treatments — not just manufacture existing ones — is significant. Indian pharmaceutical companies and research institutions that integrate AI into their drug discovery pipelines now are positioning themselves for a role in the next generation of medicine.
