SANPolyA

SANPolyA identifies polyadenylation signals (PAS) in human and mouse genomes using deep neural networks to support analysis of gene regulation and transcriptional control.

Key Features:

• Deep neural network-based methodology: SANPolyA learns complex sequence patterns directly from genomic sequences using deep neural network architectures.
• No requirement for manually crafted features: The model operates without manually engineered sequence features, relying on learned representations.
• Benchmark performance: Comparative analyses on multiple benchmark datasets indicate SANPolyA outperforms several state-of-the-art PAS identification methods in accuracy and reliability.
• Robust evaluation: The method was validated using a leave-one-motif-out evaluation strategy to assess generalizability across different PAS motifs.

Scientific Applications:

• Gene expression studies: Precise identification of PAS sites aids investigation of polyadenylation effects on mRNA stability and translation.
• Comparative genomics: Detection of PAS motifs across species supports analyses of evolutionary conservation and divergence between human and mouse.
• Disease research: Characterizing aberrant PAS recognition can contribute to understanding mechanisms of diseases involving dysregulated gene expression.

Methodology:

SANPolyA employs deep neural network architectures trained on human and mouse genomic sequences without manually crafted features, and it was evaluated via comparative benchmarking on multiple datasets including a leave-one-motif-out strategy.