LAX Leader

USC researchers have developed a new artificial intelligence model, published in Nature Methods, that predicts the accuracy of protein-DNA binding across various protein types. This advancement could significantly reduce the time required to develop new drugs and medical treatments.

The tool, named Deep Predictor of Binding Specificity (DeepPBS), is a geometric deep learning model designed to predict protein-DNA binding specificity from complex structures. Researchers can input data into an online computational tool to analyze these structures.

"Structures of protein–DNA complexes contain proteins usually bound to a single DNA sequence. For understanding gene regulation, it is important to access the binding specificity of a protein to any DNA sequence or genome region," said Remo Rohs, professor and founding chair in the Department of Quantitative and Computational Biology at USC Dornsife College of Letters, Arts and Sciences. "DeepPBS is an AI tool that replaces high-throughput sequencing or structural biology experiments for revealing protein-DNA binding specificity."

DeepPBS utilizes a geometric deep learning model to capture chemical properties and geometric contexts of protein-DNA interactions. It produces spatial graphs illustrating protein structure and the relationship between protein and DNA representations. Unlike many existing methods limited to one family of proteins, DeepPBS can predict binding specificity across various families.

"It is important for researchers to have a method available that works universally for all proteins and is not restricted to a well-studied protein family. This approach also allows us to design new proteins," Rohs added.

The field of protein-structure prediction has seen rapid advancements with tools like DeepMind’s AlphaFold predicting structures from sequences. These tools have increased the availability of structural data for analysis. DeepPBS complements structure prediction methods by predicting specificity for proteins without available experimental structures.

Rohs noted numerous applications for DeepPBS, including accelerating drug design and treatments for specific cancer cell mutations, synthetic biology discoveries, and RNA research applications.

The study's authors include Raktim Mitra, Jinsen Li, Yibei Jiang, Ari Cohen, Tsu-Pei Chiu (all from USC), Jared Sagendorf (University of California, San Francisco), and Cameron Glasscock (University of Washington). The research was primarily supported by NIH grant R35GM130376.

###