single-cell-foundation-model-langcell

What problem does it solve?

LangCell provides a language-aware framework for single-cell identity understanding, enabling zero-shot and few-shot cell type annotation guided by text descriptions and ontology context.

Core Features & Use Cases

• Zero-shot and few-shot cell type annotation by aligning cell embeddings with textual descriptions.
• LangCell-CE finetuning for cell encoder training.
• Preprocessing AnnData into the tokenized format expected by LangCell.
• Preparing text descriptions for candidate cell identities and enabling multimodal cell-text matching workflows.
• Guardrails and practical checks via local usage guidance.

Quick Start

Confirm whether the user wants zero-shot annotation, few-shot annotation, or cell-encoder-only finetuning.