A digital revolution in drug discovery – Meet Google’s protein folding AI

I remember attending a lecture several years ago that focused primarily on the 3D structures of proteins and their importance in drug discovery. It being both boring and interesting at the same time, I think many students can relate, the lecturer used his last 30 seconds of the lecture to talk about an upcoming AI that can predict how a certain protein will fold. Completely blown away by what I heard, I was impressed, and made to realize just how far technology has come. But why?

Let’s talk a bit about biology to understand, most of us learned in high school that a protein is nothing more than a chain of amino acids, which is correct. However, a chain of amino acids is not going to do much unless it is folded in a very specific structure. Proteins differ in length, generally between 50 and 2000 amino acids ¹, so try to image how many folded structures you can predict. I think a lot, but most of them wouldn’t matter, because there is only one right.

What is AlphaFold

Google has a division with a lot of big minded people, called DeepMind. After beating the world champion of chess with their AI, they now turned their attention to biology. The result was AlphaFold, an AI system that is able to predict protein structures with very good accuracy. They even solved a protein folding problem with a accuracy that was never seen in the field This AI can predict 3D shapes of proteins from their amino acid sequences in a matter of hours, a task that can take several years.

By achieving such accurate predictions, AlphaFold has shown it can be a game-changer for drug discovery.

Why is it so important?

It is all fun and impressive, but why is it so important to know the 3D structure of a protein? Well, besides serving as structural supports, hormones, building block etc ², it also serves as receptors. And receptors in the body have a lot of important functions, like signal reception and -transduction, cell communication etc. In order to ‘activate’ a receptor, a ligand needs to bind to its binding site. And that is where the importance lies, in the ligand-receptor interaction.

For certain treatments a receptor needs be blocked or stimulated, explained in the most easy way. In most cases, a research group will try to find or synthesize a ligand with a good binding affinity with the receptor. And this becomes very much easier when they have a clear 3D model of the receptor and its binding sites.

Without boring you too much, this revolutionary tool can help researcher in accelerating their studies and finding high-affinity ligands for important treatments. Treatments that, hopefully, helps all humanity and finds cures for ‘incurable’ diseases like alzheimer or cancer.