BOLD-GPCRs: A Transformer-Powered App for Predicting Ligand Bioactivity and Mutational Effects Across Class A GPCRs

Area of Science:

• Pharmacology
• Computational Biology
• Drug Discovery

Background:

• G protein-coupled receptors (GPCRs) are crucial drug targets involved in many diseases.
• Traditional drug discovery methods struggle with GPCRs lacking extensive ligand or structural data.
• There is a need for advanced strategies to predict ligand bioactivity for the entire GPCR family, particularly understudied subtypes.

Purpose of the Study:

• To introduce BOLD-GPCRs, a deep learning framework for enhanced prediction of ligand bioactivity across class A GPCRs.
• To provide a user-friendly web interface for accessing the BOLD-GPCRs tool.
• To facilitate drug discovery for GPCRs, especially those with limited available data.

Main Methods:

• Utilized transfer learning and curated datasets of class A GPCR ligands, sequences, and mutations.
• Integrated dense neural network classifiers with transformer-based protein language models.
• Developed a web interface for BOLD-GPCRs accessibility.

Main Results:

• BOLD-GPCRs demonstrates robust predictive performance for ligand bioactivity across diverse class A GPCRs.
• The framework accurately predicts the effects of receptor mutations on ligand activity.
• Achieved strong results for both well-characterized and poorly characterized GPCR subtypes.

Conclusions:

• BOLD-GPCRs is a powerful deep learning tool for predicting ligand bioactivity and mutational effects in class A GPCRs.
• The framework significantly aids in the discovery of novel therapeutics targeting GPCRs.
• BOLD-GPCRs offers a valuable solution for exploring understudied GPCRs in drug discovery efforts.