Google’s DeepMind developed AlphaFold, the first AI-driven system that predicts the structure of proteins. Why is that important? The interaction of molecules forms the basis of biology. The protein structure, made up of chains of amino acids, determines whether and how protein molecules will interact. Remember all the talk about the COVID-19 virus’ spike protein? To develop an effective vaccine, researchers needed to understand the shape of that protein and design a new protein that would interact with the spike protein of the virus.
Before AlphaFold, researchers had no way of knowing how a chain of amino acids would fold and, therefore, built the protein and then, through laborious work, detailed the structure.
Over many years, scientists identified 200,000 protein structures. AlphaFold took the information from those available protein structures and learned to predict how a chain of amino acids would fold. AlphaFold’s predictions are not 100% accurate, but they significantly limit the possible correct results. Scientists then build the proteins predicted by AlphaFold, examine each protein’s structure, and compare it to the predicted structure. Rather than building and examining 1,000s of proteins, scientists can focus on a few dozen or fewer. This dramatically speeds up pharmaceutical research development and our understanding of how proteins interact in biology.
The introduction of AlphaFold marks the start of a more significant shift toward predictive models in biology, where discoveries are made using a more limited amount of data and in much shorter time frames. AlphaFold represents a substantial step towards a deeper understanding of cellular interactions and the delivery of more personalized medicine.
I look forward to your thoughts, so please submit your comments in this post and subscribe to my weekly newsletter, “What’s Your Take?” on DocsNetwork.com.