Researchers at the Massachusetts Institute of Technology developed the EquiBind deep-learning model, which binds molecules to proteins 1,200 times faster than existing methods in drug discovery.
Before drug development, researchers must first identify molecules capable of binding to specific target proteins. However, this process requires substantial financial and computational resources. Moreover, the development and testing of a new drug take decades, and 90% of discoveries fail, the researchers said.
According to the lead author, Hannes Stark, existing methods of binding ligands to proteins resemble \”trying to insert a key into a lock with many keyholes\”.
\”Typical models are time-consuming and evaluate each ‘fit’ before selecting the best option. By contrast, EquiBind directly predicts the exact binding pose in one step without prior knowledge of the protein’s pocket, a process known as ‘blind docking’,\” the scientist said.
According to the researchers, the model has built-in geometric reasoning that helps it study the underlying physics of molecules and make more accurate predictions when confronted with new, unseen data.
Relay Therapeutics’ Chief Data Officer Pat Walters proposed testing the system on an existing drug and protein used in treating lung cancer, leukemia, and gastrointestinal tumors. He said the algorithm successfully bound ligands to proteins, something traditional methods failed to do.
\”EquiBind offers a unique docking solution that includes both pose prediction and identification of the binding site,\” Walters says.
The researchers plan to present the algorithm at the International Conference on Machine Learning. Stark said the team intends to gather more industry feedback to improve it.
In March, the AI algorithm generated more than 40,000 variants of chemical weapons in six hours.
In November 2021, Alphabet founded a new company for drug discovery with AI.
In July, DeepMind’s AI modeled 20,000 human protein structures.
Subscribe to ForkLog AI on Telegram: ForkLog AI — all the news from the AI world!