>>12408604A pretty big deal.
It's quite easy to get a gene sequence, and so it should be easy to predict the sequence of amino acids the protein has. But how do you determine the structure of a protein experimentally?
You have to find a way to produce the protein en mass, then isolate and purify it, then crystallize it or flash freeze it, then use x-ray crystallography or other techniques to gather data to determine its structure.
This is expensive and time consuming.
Imagine you want to make a designer protein, for genetic engineering or whatever. Each iteration of protein you want to determine the structure, so you can do further computational simulations. But that's expensive and time consuming.
Imagine just being able to feed potential sequences into the computer, and the computer finds the structure overnight. Or you rent a cluster of super computers to blitz through every potential sequence you are interested in.
Reportedly Alphafold 2 achieves an accuracy comparable to experimental methods, and its about 50% better than the previous version Alphafold.