LanceOtron

LanceOtron performs peak calling on ATAC-seq, ChIP-seq, and DNase-seq assay coverage tracks to identify DNA-encoded regulatory elements across the genome.

Key Features:

• Deep learning integration: Integrates deep learning models to evaluate peak shape rather than relying solely on simple height metrics.
• Image-recognition approach: Applies image recognition capabilities of deep learning to interpret complex peak shapes in coverage tracks.
• Multifaceted enrichment measurements: Combines multiple enrichment metrics with shape information to score candidate peaks.
• Addresses statistical assumptions: Mitigates limitations of traditional statistical peak callers that reduce peak shapes to maximum height and assume specific background distributions.
• Comparative performance: Demonstrates improved selectivity and near-perfect sensitivity relative to traditional tools such as MACS2.
• Supported data types: Explicitly targets ATAC-seq, ChIP-seq, and DNase-seq datasets and operates on their assay coverage tracks.

Scientific Applications:

• Peak calling: Detection and scoring of peaks in ATAC-seq, ChIP-seq, and DNase-seq data for identification of regulatory regions.
• Regulatory element identification: Identification of DNA-encoded elements manifesting as peaks within assay coverage tracks across the genome.
• Improved signal discrimination: Enhanced discrimination between true peaks and complex background signals in genome-wide chromatin assays.

Methodology:

Integrates deep learning-based image recognition with multifaceted enrichment measurements to assess peak shape and call peaks on assay coverage tracks.