LLMs + Foundational models for Single Cell Analysis

🚀 We’re entering an exciting era where Biological Foundation Models are being integrated with Large Language Models (LLMs) — opening up powerful new ways to model cellular biology.

One standout framework is Cell2Sentence, which encodes single-cell expression profiles as sentences, enabling LLMs to learn from both biological data and text. This allows for impressive capabilities like:

🔍 Cell type annotation.

🔄 Perturbation response prediction.

🧬 Generating conditional cells.

To explore the architecture hands-on, I created a simple demo script that trains the model from scratch using a local LLM — ideal for understanding how Cell2Sentence works under the hood.

📂 GitHub Repository: Cell2Sentenece-Demo

📄 Related Paper: Read the paper