CNN6mA

CNN6mA predicts N6-methyladenine (6mA) sites in genomic DNA sequences to identify epigenetic modification locations relevant to DNA replication, repair, transcriptional regulation, and disease studies.

Key Features:

• Neural network architecture: Integrates a position-specific 1-D convolutional layer and a cross-interactive network to model sequence patterns associated with 6mA.
• Position-specific 1-D convolutional layer: Applies filters with varying window sizes across sequence positions to extract multi-level position-specific features.
• Cross-interactive network: Models relationships between nucleotide patterns across the entire inquiry sequence to capture inter-positional interactions.
• Interpretability: Generates a contribution score vector that quantifies feature importance for each prediction.
• Cross-species performance: Reported to outperform existing state-of-the-art models across various species.

Scientific Applications:

• 6mA site prediction: Identification of N6-methyladenine (6mA) sites within genomic sequences for epigenetic analyses.
• Epigenetic mechanism studies: Facilitates investigation of roles of 6mA in DNA replication, repair, transcriptional regulation, and disease pathogenesis such as cancer.
• Complement to experimental detection: Provides computational predictions that complement experimental approaches including immunoprecipitation, mass spectrometry, and capillary electrophoresis.
• Comparative analyses across species: Enables cross-species comparison of 6mA patterns and model performance.

Methodology:

Implements a neural network combining a position-specific 1-D convolutional layer with varying-window filters and a cross-interactive network and outputs a contribution score vector for interpretability.