DPProm

DPProm predicts promoters and their types across phage genomes to identify promoter regions involved in transcriptional regulation for genome annotation and regulatory studies.

Key Features:

• Two-Layer Model Architecture: DPProm implements a two-layer architecture (DPProm-1L and DPProm-2L) to separate initial promoter detection from promoter-type classification.
• DPProm-1L (First Layer): A dual-channel deep neural network ensemble that integrates multi-view features to classify DNA sequences as promoters or non-promoters.
• Sequence Feature Extraction (CNNs): Convolutional neural networks process sequence data to capture sequence patterns for promoter recognition.
• Handcrafted Features: Incorporates free energy, GC content, cumulative skew, and Z curve features to represent DNA structural and compositional properties.
• DPProm-2L (Second Layer): Uses CNNs to predict promoter types, distinguishing host-derived versus phage-specific promoters.
• Novel Data Processing Workflow: Sliding window techniques and merging sequences modules enable comprehensive genome-wide prediction across phage genomes.
• Performance and Accuracy: Experimental evaluations report improved promoter prediction accuracy and reduced false positive rates compared to state-of-the-art methods.

Scientific Applications:

• Promoter identification in phage genomes: Genome-wide detection of promoter regions to support annotation and regulatory analysis.
• Promoter origin classification: Differentiation of host-derived and phage-specific promoters to study phage gene regulation.
• Phage-host interaction studies: Inform analyses of transcriptional regulation relevant to phage-host dynamics.
• Viral evolution investigations: Enable comparative studies of promoter features across phage genomes for evolutionary insights.
• Phage-based therapeutic research: Support development of phage-based strategies by characterizing regulatory elements.

Methodology:

DPProm employs a two-layer computational approach: DPProm-1L is a dual-channel deep neural network ensemble integrating multi-view features with CNN-based sequence feature extraction and handcrafted features (free energy, GC content, cumulative skew, Z curve), genome-wide prediction uses sliding window and merging sequences modules, and DPProm-2L applies CNNs to classify promoter types as host-derived or phage-specific.