as someone who is not in this field (purely doing CS research): haven't recent advances in ML "solved" most protein structures and made x-ray crystallography obsolete?
I read the article very differently. Paul Adams is sort of a crystallographer's crystallographer who shows up to write key papers at important times. His work can generally be considered "technically correct" but also strongly in favor of maintaining the viability of crystallographers. If he writes a paper showing tiny differences (those "local" and "global" differences in the abstract) then it's really an implicit "wow, AF3 is amazingly good" and "the remaining needs for crystallography grow increasingly sparse".
(my phd is in structural biology, I've worked across multiple fields of protein structure and function, have worked with Paul in the past, and work in the field of protein structure now, although mainly on the backend infrastructure. So I have some of the credentials necessary to make the statement above.
I don't think we disagree really. There's no question that AF3 is an amazing tool and the remaining needs for crystallography grow increasingly sparse.
The parent was specifically asking if crystallography was obsolete, which in my mind is a label with a very high bar, one incompatible with crystallography's continued status as the gold standard against which AF3 is compared.
