GPCRSPACE: A New GPCR Real Expanded Library Based on Large Language Models Architecture and Positive Sample Machine...

Area of Science:

• Medicinal Chemistry
• Computational Drug Discovery
• Pharmacology

Background:

• G protein-coupled receptors (GPCRs) are critical drug targets, yet many lack selective modulators.
• Developing novel GPCR therapeutics is essential due to their role in numerous physiological processes.

Purpose of the Study:

• Introduce GPCRSPACE, a novel chemical library for GPCR drug discovery.
• Leverage a new G protein-coupled receptors large language model (GPCR LLM) architecture for enhanced compound screening.

Main Methods:

• Developed GPCR LLM using a positive sample machine learning strategy, avoiding negative sample construction.
• Built the GPCRSPACE chemical library based on the GPCR LLM.
• Evaluated GPCRSPACE for synthesizability, structural diversity, and GPCR-likeness.

Main Results:

• GPCR LLM accelerates the identification and screening of potential GPCR-interactive compounds.
• GPCRSPACE demonstrates superior synthesizability, structural diversity, and GPCR-likeness compared to existing datasets.
• The positive-sample-only approach reduces false negatives and labeling time.

Conclusions:

• GPCRSPACE represents a valuable resource for advancing GPCR drug discovery.
• The GPCR LLM architecture offers an efficient approach to exploring GPCR chemical space.
• This work addresses the need for selective modulators for under-targeted GPCRs.