Deep6mAPred

Deep6mAPred predicts DNA N6-methyladenosine (6mA) sites in plant genomes to enable investigation of epigenetic regulation, DNA repair, and developmental processes.

Key Features:

• Architecture: Combines convolutional neural networks (CNNs) and bi-directional long short-term memory networks (Bi-LSTMs) in a parallel architecture to capture local sequence motifs and long-range dependencies.
• Attention Mechanism: Incorporates an attention mechanism to weight informative sequence positions for improved 6mA site discrimination.
• Performance: Achieved an accuracy of 0.9556 on an independent rice dataset.
• Cross-Species Applicability: Validated across multiple plant species, demonstrating applicability to diverse plant genomes.

Scientific Applications:

• Gene Expression Regulation: Prediction of 6mA sites to investigate their influence on gene expression patterns.
• DNA Repair and Replication Studies: Facilitates analysis of potential roles of 6mA in DNA repair and replication processes.
• Developmental Biology: Provides insights into 6mA involvement in plant growth and developmental processes.

Methodology:

Integrates CNNs to capture local sequence patterns and Bi-LSTMs to model long-range dependencies in a parallel architecture, with an attention mechanism applied to learn sequence representations for 6mA site prediction.