iDHS-Deep

iDHS-Deep predicts DNase I hypersensitive sites (DHSs) in genomic sequences using deep learning to identify regulatory noncoding regions such as promoters, enhancers, and transcription factor-binding sites.

Key Features:

• Deep learning algorithm: Uses a deep learning–based algorithm to predict DHSs from sequence data.
• DHS vs non-DHS classification: Distinguishes DHS regions from non-DHS regions in genomic sequences.
• Cross-cell-type and developmental validation: Demonstrates predictive performance across diverse datasets derived from different cell types and developmental stages.
• High predictive accuracy: Achieves high accuracy in DHS prediction relative to traditional methods.
• Regulatory element detection: Identifies regulatory components within noncoding regions, including promoters, enhancers, and transcription factor–binding sites.
• Disease-associated loci association: Highlights DHSs that often correlate with loci associated with diseases or traits.

Scientific Applications:

• Regulatory element annotation: Annotating promoters, enhancers, and transcription factor–binding sites in noncoding genomic regions.
• Prioritization of disease-associated noncoding loci: Prioritizing noncoding loci that may be associated with diseases or traits based on chromatin accessibility predictions.
• Developmental and cell-type chromatin analysis: Comparing chromatin accessibility patterns across developmental stages and different cell types.
• Support for experimental studies of gene regulation: Guiding experimental investigation of accessible chromatin regions and regulatory element function.

Methodology:

Uses a deep learning–based algorithm to predict DHSs from genomic sequences and to distinguish DHS and non-DHS regions across datasets.