Nature magazine hypes AI automated protein folding

Post Reply
Posts in topic: 1
Posts: 14
Joined: Mon Apr 27, 2020 10:48 am


Nature magazine hypes AI automated protein folding


Post by Blogio » Thu Apr 30, 2020 1:37 pm

Anything in Nature magazine these days is secretly an advertisement for some product.

In this article, the product appears to be AlphaFold. AlphaFold is a project supported by AI pushers, Deepmind, a bunch of business types with a sense of entitlement and apparently a lot of money. They are advertising that they take money from the Goog, so perhaps that's where their clout comes from also. Their bases of operation are in London, Oxford, Cambridge, and NYU (just to throw the Ammies a bone).

These sorts of predictions of protein structure, energy, etc. have been done for years by many companies, having no need for AI. Or rather I should say having no need for pretending that AI is a magic ingredient. In the article, they mention that they aren't quite ready for production because their predictions aren't accurate to 2-3 Angstroms, which is the scale that is typical of good x-ray crystallography.

However, this misses the much larger point that crystallographic structures are created under non-physiological conditions. It is well known in the field, that the x-ray crystal structures are sometimes of limited use when determining how a protein behaves. The real proteins, when placed in solution or embedded in a membrane will frequently adopt more than one configuration, sometimes having almost no relation to the crystal structure.

The conclusion of the hype piece, "this is a watershed moment for the field" is idiotic in its simplicity. What have these blokes done for anyone?

You say this is a peer-reviewed advertisement? Yes?
word count: 261

Post Reply