AlphaFold Solved Protein Structures: So Why Haven’t We Cured Everything Yet?

In 2020, a piece of software did something structural biologists had chased for fifty years: it looked at a string of amino acids and predicted, with near-experimental accuracy, the 3D shape the protein folds into. AlphaFold didn’t just solve a hard problem. It solved the problem, the one Christian Anfinsen won a Nobel Prize for framing back in 1972. By 2024, its creators, Demis Hassabis and John Jumper, had their own Nobel Prize in Chemistry for it.

So here’s the question every bright-eyed biotech student eventually asks: if we can now predict the shape of almost any protein in seconds, why haven’t we cured cancer, Alzheimer’s, and everything else in between? It’s been six years. Where are the drugs?

The honest answer is that AlphaFold solved a genuinely enormous problem, and it happened to be sitting right next to an even bigger one.

What “solving” protein structures actually bought us?

Before AlphaFold, getting the 3D structure of a single protein could mean months or years of X-ray crystallography or cryo-EM: crystallizing a notoriously uncooperative molecule, or wrestling it into a microscope, just to get one static picture. AlphaFold and its successors have now predicted structures for over 200 million proteins, covering nearly every protein sequence known to science, in a public database anyone can query for free.

That’s not a small thing. For researchers working on a target that had never been crystallized (which used to mean starting from scratch), a decent structural hypothesis is suddenly free and instant. It’s accelerated basic research, informed which targets even look “druggable,” and started acting as a scaffold that speeds up experimental structure determination itself, since a predicted model gives crystallographers a template to solve real data faster. Prediction and experiment turned out to reinforce each other rather than compete.

Where the gap actually opens up ?

Here’s the part that doesn’t make it into the headlines: a protein structure is not a drug target. It’s the starting sketch.

AlphaFold predicts one static shape: the most probable conformation, in isolation, with no drug molecule anywhere near it. But proteins aren’t statues. They flex, breathe, and often change shape around the very molecule you’re trying to design against them, a phenomenon called induced fit. The binding pocket a drug actually needs to fit into may simply not exist in the conformation AlphaFold handed you. AlphaFold 3, released in 2024, made real progress modeling proteins together with ligands, but researchers in the field are candid that it still struggles with exactly the cases that matter most for drug design: large conformational shifts, and a tendency to default toward one familiar functional state over others.

That means every structure still needs to be treated as a hypothesis, not a blueprint. Medicinal chemists still have to confirm binding-site detail experimentally before committing months of lead-optimization work to it: the unglamorous, expensive, slow science that was always the actual bottleneck.

The number that keeps everyone honest

Structure prediction sits at the very front of a pipeline that runs: target identification, structure, hit discovery, lead optimization, preclinical safety, Phase I, Phase II, Phase III, FDA review. AlphaFold’s real contribution lives almost entirely in that first slice.

As of mid-2026, that shows up starkly in the numbers. Zero AI-discovered drugs have received full FDA approval. Out of roughly 6,100 drugs currently in clinical development worldwide, only about 1% trace back to an AI-discovered molecule, and a much smaller sliver (a few dozen) to genuinely AI-discovered targets. The furthest-along candidate, Insilico Medicine’s rentosertib for pulmonary fibrosis, only entered Phase III trials in July 2026.

The one place AI-assisted approaches do show a real edge is Phase I: AI-derived molecules are clearing early safety trials at roughly 80–90%, well above the historical 40–65% baseline, probably because these molecules are optimized on rich datasets from the start. But by Phase II, where a drug has to actually work, that advantage evaporates to roughly the historical average. Across all phases combined, AI-discovered drugs succeed at maybe 9–18%, versus 5–10% historically: better, but nowhere near the leap the “AI will cure disease” headlines implied.

So did AlphaFold actually change anything?

Yes, just not the thing most people assumed. It compressed the front end of drug discovery, in some reported cases shrinking the traditional 6–8 year discovery-to-candidate phase down to 18–30 months. It made structural biology a starting resource instead of a multi-year research project in itself. It’s reshaping which questions are even askable in academic labs with no budget for a crystallography core.

What it didn’t do, because this was never really the bottleneck it could fix, is shrink the years of safety testing, the manufacturing scale-up, the regulatory review, or the biology of why a molecule that looks perfect on a screen still fails when it meets an actual human body. That part of the pipeline is still measured in years and billions of dollars, structure prediction or not.

Why this matters if you’re headed into this field ?

If you’re a student weighing a future in computational biology versus the wet lab, the real lesson from AlphaFold isn’t “learn to code and skip the pipette.” It’s that the highest-value people in this next decade will be the ones who can move fluently between a predicted structure and the experiment that tests it, who know exactly how much to trust a model, and exactly when to go validate it at the bench. AlphaFold didn’t replace structural biology. It changed what structural biologists, and the chemists and clinicians downstream of them, spend their time on.

The cure isn’t stuck because the science stalled. It’s stuck exactly where it’s always been stuck: in the long, expensive, human part of drug development that a shape prediction, however brilliant, was never going to shortcut.

References:-

• AlphaFold in Drug Discovery: What It Has and Hasn’t Changed — Drug Discovery News
• AI Drug Discovery FDA Approvals: The 2026 Reality Check — IntuitionLabs
• Why AlphaFold won’t revolutionise drug discovery — Chemistry World