5 ms·
I wonder how accurately we can run mutation simulations updated based on best current data to correct the best guess optimization function(s). It would be a bre
by datameta 5y ago
I wonder how accurately we can run mutation simulations updated based on best current data to correct the best guess optimization function(s). It would be a breakthrough if we can predict the likeliest mutation. I'm aware of the enormous number of variables making this tremendously difficult...
- nradov 5y agoWe don't have the technology to stimulate that accurately in software. It would have to be done using transgenic animal models and/or human tissue cultures. Essentially it would be a form of gain-of-function research, and there are concerns that a lab accident could result in an even more dangerous variant escaping into the wild.
- datameta 5y agoWith ML-assisted drug discovery in, say, kinase inhibitors we search the molecule variation space for test candidates. What stops us from applying this to the protein scale by folding binding sites to discover variant candidates that would be more virulent? I'd like to understand this better.
- nradov 5y agoSure you can explore the solution space in software to identify candidates but at some point you have to create a live virus and see how it actually performs.