Paper summary: Our work on a novel protein structure shows how machine learning can also help experimental science

Lazaro et al. Communications Biology 2021 [open access here]

You may have heard last year about the hype on Deepmind’s artificial intelligence program AlphaFold and AlphaFold2 “solving” one of the most important biological problems out there: predicting the 3D structures of proteins. Well, it wasn’t truly solved, but it is true that the Alphabet(Google)-owned company gave a big step forward in the direction of some day leaving aside the complex, laborious and expensive efforts required to determine protein structures experimentally. But experiments are still needed and will be needed for a loooong time.

An important point is that these new technologies (not Deepmind’s, because it is not open, but similar ones developed by academic labs who provide them for free) can already today assist experimental procedures for structure determination. This effectively accelerates discoveries, enables better use of the data produced, and overall decreases time and money spent.

In this open access paper my friend Natalia led an effort to experimentally determine the first structure of a large glutamate dehydrogenase enzyme. Part of the work was assisted by models of the enzyme predicted ab initio using the most modern tools that exploit contact and distance predictions, which I have discussed in this other review.

Keywords: English, protein, molecular modeling, deep learning, machine learning, artificial intelligence, casp, structure prediction, rosetta, raptorx, CASP

I am a nature, science, technology, programming, and DIY enthusiast. Biotechnologist and chemist, in the wet lab and in computers. I write about everything that lies within my broad sphere of interests. Check out my lists for more stories. Become a Medium member to access all stories by me and other writers, and subscribe to get my new stories by email (original affiliate links of the platform).