RAGPPI: RAG Benchmark for Protein-Protein Interactions in Drug Discovery

The article introduces the RAG Benchmark for Protein-Protein Interactions (RAGPPI), a new benchmark consisting of 4,420 question-answer pairs designed to evaluate the biological impacts of protein-protein interactions in drug discovery. It includes a gold-standard dataset of 500 QA pairs, annotated by experts, and a silver-standard dataset of 3,720 QA pairs, assessed using an ensemble auto-evaluation LLM based on expert labeling and similarity metrics. This benchmark is significant for practitioners as it provides a structured resource to enhance Retrieval-Augmented Generation systems, facilitating more efficient target identification in drug development.