The AlphaFold Mania
The new protein structure solving AI from Google is shaking up biology
Alphafold from Deepmind is an earthquake that keeps shaking the world of biology.
If you have any doubt, take a look at the image above that shows the number of articles published citing Alphafold2.
The paper describing AlphaFold2 and its source code was released in July 2021. That’s also where the graph begins to swerve upwards.
For those in the dark, AF2 is an AI library from DeepMind that predicts the 3D structure of a protein starting from only the base amino acid sequence.
This was a particularly notorious problem in biology — that was tackled through heuristics and first principles-based computer code until DeepMind came along.
They showed that rather than ‘us’ coding what each amino acid had to do and how we could train an AI to learn the rules by observing hundreds of examples.
And the AI defeated us all.
Today, AlphaFold is the basis of most of the work in proteomics and drug discovery.
Each drug interacts with some protein or the other. Knowing the structure of the protein goes a significant step in determining how a particular drug/cancer/gene etc interacts with it.
