Depicter

Depicter identifies and classifies eukaryotic promoter regions near transcription start sites (TSS) to support analysis of transcription initiation and transcriptional regulation.

Key Features:

• Deep Learning Architecture: Employs a cascaded deep capsule neural network architecture that integrates convolutional neural networks with capsule layers for promoter prediction.
• Promoter Classification: Distinguishes three promoter types: TATA-box-containing (TATA), TATA-box-lacking (non-TATA), and a combined TATA/non-TATA class for indistinguishable promoters.
• Species-Specific Dataset: Trained on a curated, species-specific promoter dataset including Homo sapiens, Mus musculus, Drosophila melanogaster, and Arabidopsis thaliana.
• Biological Targeting: Targets promoter regions proximal to TSS that mediate RNA polymerase recruitment and transcription initiation.
• Performance: Validated by benchmarking and independent testing with reported superior predictive performance relative to several state-of-the-art promoter identification methods.

Scientific Applications:

• Promoter annotation: Identification and classification of promoter locations and types to inform genomic studies and gene regulation analyses.
• Transcriptional regulation studies: Investigation of mechanisms of transcription initiation by locating promoter regions involved in RNA polymerase recruitment.
• Cross-species analysis: Comparative analysis across Homo sapiens, Mus musculus, Drosophila melanogaster, and Arabidopsis thaliana to explore conserved and divergent promoter features.
• Gene expression analysis: Providing promoter-type information that can be correlated with expression patterns and regulatory outcomes.
• Genetic engineering: Informing selection and design of promoter elements for construct development based on predicted promoter types.

Methodology:

Models are trained using a cascaded deep capsule neural network combining convolutional neural networks with capsule layers on a curated promoter dataset, with separate model variants for TATA, non-TATA, and combined classification, and evaluated via benchmarking and independent testing.