A major breakthrough in biomedical science?

Maybe. For some reason this has flown under the radar and I haven't seen it reported in the lay media. Basically, the claim is that Google AI has solved the protein folding problem, at least well enough to be useful. Before I explain what this means, to put it in context, a central fact about medicine is that it is continually advancing technically. You may think this is an unalloyed good but the reality over pretty much the past century is that there's a downside. 

 

Social factors are still more important than medical intervention in determining population health -- clean air and water, access to high quality affordable nutrition, social support, good quality housing, all that sort of stuff. But we don't invest nearly enough in social welfare while medicine soaks up more and more of the economy -- now almost 18%, and it keeps growing. All of this high tech stuff means there are more and more situations in which medical intervention is possible, and it's also very expensive. Sure, you'd probably rather than spend the money than die or be in pain and disabled, but the point is we could better spend some of the money elsewhere. That's a complicated subject, but it's my core purpose here so I thought I should reiterate it.

So, protein folding. To review what you already know, proteins are what genes code for. They are long chains of smaller molecules called amino acids, of which there are 20 that are important in living organisms. Ribosomes are structures in the cell that build proteins from the messenger RNA that conveys information from the DNA in the nucleus. The proteins come off the ribosome as linear chains but they immediately fold into complex structures, determined by the ways in which amino acids in the chain attract others. These complex structures determine how the proteins function, notably including how they interact with each other. For example, Covid-19 gets its genome into cells when the notorious spike protein binds with the ACE receptor on the cell membrane, which is also a protein structure.

However, determining the structure of proteins has been extremely difficult. To fully describe a single protein has taken many years of laboratory work. But now we can sequence DNA, indeed the entire human genome has been sequenced, which means that the amino acid sequence of proteins is immediately knowable. However, protein folding is such a complicated process that it was not possible to predict the structure of a protein from its linear amino acid sequence. Now, as of this year, it is, thanks to immense advances in computational power and machine learning techniques. This is an impressive example of the synergy of science -- technological advances and discoveries in seemingly unrelated fields combine to make further advances possible.

Now, I'm not going to get too carried away. Everybody thought that sequencing the genome in itself would make a powerful impact on medical practice, but so far the impact hasn't been all that dramatic. This could change that conclusion, however. But it remains to be seen whether the next step, of really being able to predict how proteins will interact with each other and therapeutic targets can be found, will come any time soon. But the imaginable possibilities are mind boggling.