Deep-4mCW2V

Deep-4mCW2V predicts N4-methylcytosine (4mC) modification sites in Escherichia coli genomic sequences to support analysis of DNA methylation patterns relevant to transcription regulation, replication, and gene expression.

Key Features:

• Deep learning-based model: Uses a deep learning framework to predict 4mC modification sites in Escherichia coli genomic sequences.
• Word embedding (word2vec): Encodes DNA sequences with the word2vec technique to capture contextual relationships within sequence data.
• 1-D convolutional neural network (1-D CNN): Processes word2vec-encoded features with a one-dimensional CNN to discriminate between 4mC and non-4mC sites.
• Performance: Reports an overall accuracy of 0.861 on independent datasets and outperforms existing models by approximately 4.3%.

Scientific Applications:

• DNA methylation mapping: Enables identification of 4mC sites for mapping DNA methylation patterns in Escherichia coli.
• Gene regulation and expression studies: Supports investigation of 4mC roles in transcription regulation, replication, and gene expression.
• Microbial genetics and epigenetics: Provides site-level 4mC predictions to aid studies in microbial genetics and epigenetics.

Methodology:

DNA sequences are encoded using word2vec and the resulting embeddings are input into a 1-D CNN deep learning model to distinguish modified (4mC) from unmodified sites.