What problem does it solve?
This Skill simplifies complex machine learning tasks on genomic interval data (like BED files), enabling advanced analysis of DNA regions and single-cell data without deep ML expertise.
Core Features & Use Cases
- Genomic Embeddings: Learn vector representations of genomic regions (Region2Vec) or single cells (scEmbed) for similarity analysis and clustering.
- Metadata Integration: Jointly embed regions and their associated metadata (e.g., cell types) for powerful cross-modal searches (BEDspace).
- Reference Building: Create standardized "universe" peak sets from multiple experiments for consistent analysis.
- Use Case: Analyze single-cell ATAC-seq data to identify distinct cell populations by generating cell embeddings and clustering them using scanpy.
Quick Start
Use the geniml skill to train Region2Vec embeddings on BED files in the 'bed_files/' directory, saving the model to 'model/'.