DeepArk

DeepArk predicts context-specific regulatory features from genomic DNA sequences across the model species Caenorhabditis elegans, Danio rerio (zebrafish), Drosophila melanogaster (fruit fly), and Mus musculus (mouse), modeling chromatin states, histone modifications, transcription factor binding sites, and other regulatory elements.

Key Features:

• Model architecture: Implements a suite of deep convolutional neural networks for sequence-based regulatory prediction.
• Regulatory feature prediction: Predicts chromatin states, histone modifications, transcription factor binding sites, and other regulatory elements directly from genomic sequences.
• Cross-species scope: Generates predictions for Caenorhabditis elegans, Danio rerio (zebrafish), Drosophila melanogaster (fruit fly), and Mus musculus (mouse).
• High-throughput activity profiling: Predicts thousands of distinct context-specific regulatory activities across the supported species.
• Variant impact assessment: Evaluates the potential regulatory consequences of genomic variants, including common, rare, and previously unobserved variants.
• In silico saturated mutagenesis: Profiles regulatory potential of sequences by systematically mutating positions to assess effects on predicted regulatory activity.
• Validation: Predictions have been validated against in vivo experimental studies.

Scientific Applications:

• Regulatory annotation: Annotates regulatory elements and chromatin states across multiple model organisms for comparative regulatory genomics.
• Variant interpretation: Prioritizes and interprets the regulatory impact of common, rare, and novel genomic variants.
• Mutation effect mapping: Uses in silico saturated mutagenesis to identify sequence positions critical for regulatory activity.
• Cross-species transcriptional regulation studies: Facilitates investigation of conserved and species-specific transcriptional regulatory mechanisms.
• Annotation of understudied species: Expands regulatory annotation capabilities to less-characterized model organisms within the supported set.

Methodology:

Implements a suite of deep convolutional neural networks that predict context-specific regulatory features directly from genomic DNA sequences.