scembed-atac-annotation

What problem does it solve?

scEmbed enables fast cell-type annotation for scATAC-seq data by transferring labels from a reference using region-based embeddings and a vector database.

Core Features & Use Cases

• Region-based embeddings: encode regulatory regions first and average to create cell representations for transfer learning.
• Model-consistency enforcement: ensure query uses the exact same pre-trained model as the reference to maintain a coherent embedding space.
• Fast, scalable annotation: perform KNN-based label transfer against a reference DB to annotate thousands of cells in seconds.
• Use cases: map query scATAC cells to published cell-type repertoires and rapidly re-annotate when references are updated.

Quick Start

Encode your query data with the same pre-trained databio model used for the reference and perform a KNN-based annotation against your vector store.