Google AI Researchers Say Their Model Can Predict the Structure of Proteins, a Major Scientific Goal

**Google AI Researchers Say Their Model Can Predict the Structure of Proteins, a Major Scientific Goal**.

* [Google AI blog post](https://blog.google/technology/ai/alphafold-protein-structure-prediction/).

* [Nature paper](https://www.nature.com/articles/s41586-021-03828-1).

Predicting the structure of proteins is a major scientific goal, with applications in drug discovery, materials science, and more. But it’s a notoriously difficult problem, and current methods are often inaccurate or computationally expensive..

Now, researchers at Google AI say they have developed a new model that can predict the structure of proteins with unprecedented accuracy. The model, called AlphaFold2, was trained on a dataset of over 100,000 protein structures, and it can now predict the structure of a protein in just a few minutes..

In a paper published in Nature, the researchers showed that AlphaFold2 outperforms all other existing methods on a variety of benchmarks. For example, on a set of 25 proteins that have been particularly difficult to predict, AlphaFold2 was able to predict the structure of 23 of them to within 1 angstrom (Å) of the experimental structure. This is a significant improvement over previous methods, which typically could only predict the structure of proteins to within 2-3 Å..

The researchers say that AlphaFold2 could have a major impact on a wide range of scientific fields. For example, it could be used to design new drugs, develop new materials, and understand how proteins function..