← back
From Tokens to Cells: Foundation Models for Single-Cell Biology - Akram Baharlouei, Altos Labs
Takeaway
Single-cell AI depends on handling noisy biological measurements and missing modalities, beyond scaling familiar transformer recipes.
Summary
- Single-cell foundation models aim to represent cellular state and responses, with eventual applications in rejuvenation and more efficient drug development.
- RNA sequencing dominates available training data because it scales more readily than proteomics, while imaging and other modalities capture complementary information.
- Measurements are noisy snapshots of dynamic biological processes, with additional variation from laboratories and instruments; more samples alone do not resolve this heterogeneity.
- Transformer approaches such as scGPT represent cells using gene tokens and learn through masked prediction, borrowing techniques from language modeling.
single-cellbiological-modelstranscriptomics
Original description
This talk examines the engineering challenges of building foundation models for single-cell biology from a non-biologist’s perspective. Speakers: Akram Baharlouei (Altos Labs): Machine learning engineer at Altos Labs working on foundation models for biology. Previously at Meta AI and Qualcomm. LinkedIn: / akram-baharlouei-61784421