Deepm6A-MT

Deepm6A-MT predicts RNA N6-methyladenosine (m6A) modification sites in eukaryotic mRNAs across multiple tissues and species to enable studies of epitranscriptomic regulation.

Key Features:

• Advanced Deep Learning Architecture: The model integrates bidirectional gated recurrent units (Bi-GRU) with convolutional neural networks (CNN) to capture complex sequence patterns.
• Dual Input Channels: One channel uses an embedding layer followed by Bi-GRU and CNN layers, while the second channel uses one-hot encoding, dinucleotide one-hot encoding, and nucleotide chemical property codes.
• Robust Validation: Performance was evaluated using 5-fold cross-validation and independent testing.
• Cross-Species and Cross-Tissue Testing: The model has been assessed across different species and tissue types to test generalizability.

Scientific Applications:

• Gene expression regulation: Investigating the role of m6A modifications in the regulation of gene expression.
• Tissue-specific regulation analysis: Exploring tissue-specific regulatory mechanisms involving m6A.
• Comparative epitranscriptomics: Conducting cross-species comparative studies to examine evolutionary aspects of RNA modification.

Methodology:

Two-channel model architecture with an embedding layer plus Bi-GRU and CNN in one channel and one-hot, dinucleotide one-hot, and nucleotide chemical property encodings in the other; validated by 5-fold cross-validation and independent testing and evaluated across species and tissues.