Predicting a protein's shape from its sequence is an unsolved problem

What actually happened

The "protein folding problem" had a clear statement and a scoreboard. It was posed in the 1960s, and from 1994 the biennial CASP contest measured how badly computers predicted structures from sequence. Progress was real but glacial, and the honest expectation was that a general solution lay a long way off.

In 2020, AlphaFold2 won CASP14 so decisively that the organizers declared the problem essentially solved for most practical purposes. A task that used to mean months or years of X-ray crystallography for a single protein now takes minutes by looking it up. The predictions are not perfect and do not replace experiment, but they are good enough to be the starting point for most structural biology on Earth.

It is worth being precise about what cracked. AlphaFold predicts the folded shape; it does not fully explain the physical process of folding, which is a separate question still under study. But the specific claim your textbook made, that we could not read a protein's structure off its sequence, went from true to false inside about a year, and a Nobel Prize followed.