What changed for researchers

DeepMind and the European Bioinformatics Institute put the predictions in a free public database. By late 2023 it covered more than 214 million protein sequences.

The practical change is narrower than the headlines suggested, and more useful than they made it sound.

A researcher can now start by inspecting a predicted structure on a screen. That may help choose an experiment, compare proteins, or identify a region that deserves closer study.

That is what the tool actually does. It moves the starting line.

It does not move the finish line. A prediction is a hypothesis about a shape. Researchers still verify the ones that matter, and predictions are more reliable for some kinds of protein than others. The database reports a confidence score for every region of every structure precisely because parts of it should not be trusted equally.

One slow step in a long process became fast. The prediction does not remove the experiments needed to answer later biological questions.