Folding@home ?
wasn't this one of the earlier distributed computing projects ? Along with SETI@home ?
AI can help scientists spot tiny folding protein crystals, and thus one day potentially assist eggheads in designing new drugs, according to a paper published in PLOS One. To demonstrate this form of boffinry is possible, a large team of researchers from academia and industry, including bods at Duke University in the US and …
Folding@home tackles a different beast: protein folding itself.
This article describes a method to aid in the other type of grunt work: the crystallisation of proteins into stable crystals that can be used for structural research. Spotting the things once and *if* they form can be a bit of a biatch.
This was a completed project on the World Community Grid, basically stopped because the funding was cut.
As well as detecting crystals they were trying to work out the the best conditions for getting the proteins to crystallize, which was a big multi-variate problem.
https://www.worldcommunitygrid.org/research/hcc1/overview.do
It's an achievement but I would say incremental rather than revolutionary. There's a dichotomy in modern science: think about a problem, or throw machinery at it. High energy physics and genetics are in the same part of the cycle at the moment: upgrading the LHC, and the vast sequencing projects are exemplary. We don't really understand the conditions needed for predictable manufacture of these analysable structural assemblies of proteins, so we have to use shotgun approaches. The other big thrust in this area currently is cryo-electron microscopy, where you take hundreds of thousands of images of your tiny blobs and use IT to reconstruct their common features into a structure. These approaches give you a quick return on the "low hanging fruit" but once that is harvested you are reduced to trying increasingly wide screens in the hope of landing on the low probability jackpot. As the article (and the paper) note, this is only one part of the structure determination process: making the proteins the right way in the first place is an as yet unsolved problem. Whether that end is susceptible to AI approaches, I don't know.