DeeReCT-TSS

DeeReCT-TSS identifies transcription start sites (TSSs) genome-wide by integrating DNA sequences and RNA-seq data with deep learning to provide precise TSS annotation for studies of gene regulation.

Key Features:

• Deep learning approach: Employs a deep learning framework that integrates sequence and expression features to predict TSSs genome-wide.
• Integration of multi-source data: Combines genomic DNA sequences and conventional RNA-seq data to improve TSS detection and reflect transcriptional activity across cell types.
• Meta-learning-based extension: Implements a meta-learning extension to enable simultaneous annotation across multiple cell types and to identify cell-type-specific TSSs.
• Validation and precision: Validated against independent ENCODE datasets by correlating predicted TSSs with experimentally defined TSS chromatin states, demonstrating high precision.

Scientific Applications:

• Genome-wide TSS annotation: Generates precise annotations of TSSs across the genome to support studies of promoter architecture and transcriptional regulation.
• Cell-type-specific TSS discovery: Identifies cell-type-specific TSSs to investigate differential promoter usage during cellular differentiation and disease.
• Integration with RNA-seq analyses: Links sequence-based TSS predictions with RNA-seq evidence to study transcription initiation and expression dynamics.

Methodology:

A deep learning model trained on DNA sequences and RNA-seq data, a meta-learning-based extension for multi-cell-type annotation, and validation by comparison to ENCODE-defined TSS chromatin states.