PhaSepDB 3.0: a comprehensive knowledgebase of phase separation-related proteins from AI-assisted curation
Kaiqiang You, R. Li, Ruixin Lian, Yuxuan Li, Hongzhining Yang, Yiran Zhou, Yangsheng Chen, Likun Wang, Zhaoqing Fan, Liantao Ma, Tingting Li
Abstract
Phase separation (PS) is a fundamental principle driving the formation of membraneless organelles (MLOs), which are critical for various cellular functions and pathological conditions. We present PhaSepDB 3.0 (https://db.phasep.pro/), a significantly updated knowledgebase of proteins related to PS. To address the challenges of curating a vast body of literature, we have implemented a novel human-AI collaborative workflow that integrates a large language model (LLM)-based agentic system with expert verification, enabling a major expansion and enrichment of the database. PhaSepDB 3.0 now contains 3,484 expert-curated entries for 1849 PS-related proteins, more than doubling the content of the previous version. The annotation framework has been restructured to capture deeper insights, including functional relevance, experimental evidence, and the intrinsic and extrinsic regulations of PS. A key new feature is the protein-wise summary page, which synthesizes data from multiple publications to provide a comprehensive overview of each protein's PS behaviour and functional relevance. With redesigned, user-friendly web interfaces, PhaSepDB 3.0 serves as a critical resource for the community, supporting researchers to explore the intricate basis of PS and its biological implications in greater detail.