After the Fold: Biology's Prediction Problem Moves On

A few years ago, predicting a protein's folded shape from its sequence was a grand challenge. Today it is, for most proteins, a solved problem — one of the cleanest victories machine learning has handed biology.

The aftermath is instructive. Knowing a structure is not the same as knowing a function, and proteins rarely sit still: they flex, bind, and change shape in ways a single predicted snapshot cannot capture.

Attention has shifted to the harder questions — interactions, dynamics, and the disordered regions that defy a fixed structure entirely. The map was drawn; now we must learn to read the territory's motion.

It is a familiar pattern in science: solving the question everyone asked reveals the deeper questions no one knew to.