treearches-hierarchy-learning

What problem does it solve?

Harmonizes disparate cell-type annotations across multiple scRNA-seq reference atlases by learning and applying a unified hierarchical taxonomy, enabling consistent cross-study analysis and detection of novel cell types.

Core Features & Use Cases

• Learn a unified cell-type hierarchy across studies using treeArches and scHPL on latent-space representations.
• Detect novel cell types not present in references and update hierarchies with new labeled data.
• Map new query datasets to the learned hierarchy for consistent annotation across cohorts and conditions.

Quick Start

Provide multiple annotated references to treeArches, then run the hierarchy learning and query mapping steps to generate an updated tree and per-node classifiers.