Skip to content

How AI Revolutionized Protein Science, but Didn’t End It | Quanta Magazine

How AI Revolutionized Protein Science, but Didn’t End It | Quanta Magazine
Three years ago, Google’s AlphaFold pulled off the biggest artificial intelligence breakthrough in science to date, accelerating molecular research and kindling deep questions about why we do science.

Table of Contents

In December 2020, the protein folding problem, which involves predicting the three-dimensional shape of a protein molecule from its one-dimensional molecular code, was solved by artificial intelligence. The breakthrough was presented at a conference by John Jumper, a relative newcomer to the protein science community, who unveiled AlphaFold2, an artificial intelligence tool developed by Google DeepMind. The tool's predictive models of 3D protein structures were over 90% accurate, marking a significant advancement in the field of protein science.

The success of AlphaFold2 has revolutionized the study of proteins and inspired new algorithms, biotech companies, and ways to practice science. However, while AlphaFold2 is a powerful prediction tool, it has not replaced biological experiments but emphasized the need for them. Despite its impact, there are still significant gaps that artificial intelligence has not filled, such as simulating how proteins change through time or modeling them within cells.

The protein folding problem has been a long-standing challenge in the scientific community, dating back to the 1930s. Experimentalists and computational scientists have worked tirelessly to understand how proteins fold into their innate shapes. The Critical Assessment of Structure Prediction (CASP) experiment, founded by John Moult and Krzysztof Fidelis, provided a proving ground for computational approaches to the protein folding problem. Through this process, clear leaders emerged, such as David Baker, who developed the high-performing algorithm Rosetta.

The breakthrough achieved by AlphaFold2 and the ongoing efforts of scientists have reshaped the future of artificial intelligence in biology and have paved the way for new advancements in protein science. Despite the progress, the protein folding problem continues to present challenges, and the need for both experimental and computational approaches remains essential.