Post 23: One year after AlphaFold2 📅
Published:
AlphaFold2 (AF2) was announced about ~2 years ago. Its publication just about ~1 year ago is the most famous article of its kind (that I come across), as approximately 817,000 people have accessed it to read, and 5,703 people have cited it. In just one year!
Since then, many other AIs for predicting protein structure have emerged: OpenFold, OmegaFold, UniFold, FastFold, ColabFold, DeepFold, ESM-IF1, RGN2, etc. However, in my experience using them, I can say that AF2 still provides the best results, besides being the AI that has considered the most biological details in its training. We don’t know how much the development of AF2 cost, but its training alone is estimated to range from $25,000 to $200,000.
We are a few months away from the CASP15 (December 10-13), and it will be very exciting to see what the results will be. Here are some interesting links regarding AF2:
Refs:
Mini documentary of AF2 by DeepMind:
The best video I’ve seen to understand AF2 in detail:
Stories of how scientists have solved problems with AF2:
A list of other AIs to do protein analysis