CPRiL

CPRiL extracts functional relationships between small molecules and proteins from PubMed to support analysis of compound–protein interactions.

Key Features:

• Comprehensive Database: CPRiL encompasses approximately 465,000 unique names and synonyms for small molecules, about 100,000 distinct proteins, and catalogs over nine million functional relationships between these entities.
• Automated Extraction: CPRiL uses BioBERT machine learning models to automatically extract and determine functional relationships from textual data in biomedical articles.
• Performance Metrics: CPRiL achieved benchmark metrics of F1 = 84.3%, precision = 82.9%, and recall = 85.7% on evaluation datasets.

Scientific Applications:

• Drug discovery: Provides evidence of small molecule–protein interactions to inform the development of new therapeutic strategies.
• Molecular biology research: Aids investigation of how small molecules affect cellular functions and tissue responses.
• Metabolism and pathway analysis: Supports elucidation of impacts on human metabolism and metabolic pathways by linking compounds to protein functions.

Methodology:

CPRiL applies BioBERT models trained and tested on biomedical text to automatically extract and determine functional relationships between small molecules and proteins from PubMed articles.