Thirty-year drug company veteran Derek Lowe slams AlphaFold: "Pure self-importance" to make drugs based on structural predictions

But in the past year or two, AlphaFold has made remarkable progress in this area of work, predicting the organizational structure of the vast majority of human proteins.

A decade ago, such results would have looked like a science fiction story.

Thirty-year drug company veteran Derek Lowe slams AlphaFold: media exaggeration, relying on structure prediction for drugs is "pure self-importance

I don't want to deny the results of AlphaFold, but some news reports have misunderstood the significance of AlphaFold's results.

We have not made a huge leap in understanding "why proteins fold the way they do". Protein structures often exist as coils, rings, or sheets, but why not go deeper into them?

There are many hidden answers that would not be discovered if the research were done only at the current level.

We have long had thousands of predictions of new protein structures, the vast majority of which are correct. And, despite a few exceptions, they do seem to be mostly correct.

AlphaFold's algorithm will not work properly in the face of disordered protein regions, and AlphaFold's entire computational technique is based on finding analogues of known structures, and there is nothing AlphaFold can do in the absence of comparable structures.

A fraction of disordered proteins are able to make ordered arrangements under the influence of various proteins, but there is also a fraction of proteins that never appear to have an ordered structure under any conditions.

When proteins fail to form an ordered structure, it is beyond the computational power of AlphaFold.

AlphaFold provides confidence levels for its structure predictions. The dark blue structure has a higher confidence level, while the yellow and orange structures have a lower confidence level

It is important to emphasize that with AlphaFold, we are getting a prediction of the protein structure, not the real protein structure.

AlphaFold is a very practical method for protein prediction, but the only way to determine its accuracy is to obtain the actual data of the protein by X-ray, NMR or cryoelectron microscopy.