4mC-w2vec

4mC-w2vec identifies DNA N4-methylcytosine (4mC) sites in Rosaceae genomes using distributed sequence representations and a double-layer convolutional neural network for epigenetic site prediction.

Key Features:

• Automatic Feature Selection: Employs automatic feature selection via distributed feature representation rather than hand-crafted sequence features.
• Distributed Representation and Word Embedding: Represents sequences using k-mer sub-word units and transforms them into vector features using the word2vec embedding technique.
• Deep Learning Model: Uses a double-layer convolutional neural network (CNN) to classify sequences for the presence or absence of 4mC sites by learning complex sequence patterns.
• Versatility Across Datasets: Applicable to both balanced and imbalanced datasets for 4mC site prediction.

Scientific Applications:

• Rosaceae genomes (Rosa chinensis and Fragaria vesca): Applied to identify 4mC sites in the genomes of Rosa chinensis and Fragaria vesca within the Rosaceae family.
• Epigenetic analysis of gene regulation and transcriptional control: Supports studies of how DNA N4-methylcytosine (4mC) influences gene regulation and transcriptional control.

Methodology:

Sequences are processed to extract k-mers, k-mers are embedded using word2vec to obtain distributed feature vectors, and the embedded features are input to a double-layer convolutional neural network (CNN) for classification of 4mC sites.