Contextual AI Models for Single-cell Protein Biology

Explore a 32-minute conference talk on contextual AI models for single-cell protein biology presented by Michelle Li from Valence Labs. Dive into the introduction of PINNACLE, a geometric deep learning approach that generates context-aware protein representations. Learn how this innovative method leverages multi-organ single-cell atlas data to produce protein representations across various cell type contexts and tissues. Discover how PINNACLE's embedding space reflects cellular and tissue organization, enabling zero-shot retrieval of tissue hierarchy. Understand the application of pretrained protein representations in downstream tasks, including enhancing 3D structure-based representations for immuno-oncological protein interactions and investigating drug effects across cell types. Explore PINNACLE's superior performance in nominating therapeutic targets for rheumatoid arthritis and inflammatory bowel diseases, and its ability to identify cell type contexts with higher predictive capability than context-free models. Gain insights into the mathematical formalization, setup, and results of this groundbreaking approach. The talk concludes with a focus on RA-PINNACLE and a Q&A session, providing a comprehensive overview of this cutting-edge research in AI-driven protein biology.

- Intro + Background
- PINNACLE
- Mathematical Formalization
- Setup
- Results
- RA-PINNACLE
- Q&A