MSNet-4mC

MSNet-4mC identifies DNA N4-methylcytosine (4mC) sites in nucleotide sequences using a neural network to provide accurate computational detection of this epigenetic modification.

Key Features:

• Multi-Scale Representation Learning: Employs a lightweight neural network with convolutional operations across multi-scale receptive fields to capture short- and long-range interactions within DNA sequences.
• Class Imbalance Handling: Incorporates class weights into the cross-entropy loss during training to mitigate species-specific imbalance among candidate 4mC sites.
• Performance and Benchmarking: Extensive benchmarking experiments demonstrate superior performance compared with existing state-of-the-art methods for 4mC site identification.

Scientific Applications:

• 4mC Site Identification: Computational detection of DNA N4-methylcytosine sites from sequence data.
• Genome-scale Epigenetic Analysis: Genome-wide epigenetic profiling and analysis in genomic studies requiring rapid and accurate 4mC detection.
• Regulatory Role Investigation: Support for studies in genetics, molecular biology, and bioinformatics aimed at elucidating the regulatory roles of 4mC.

Methodology:

MSNet-4mC uses a convolutional neural network (CNN) tailored for DNA sequence analysis, employing convolutional operations across multi-scale receptive fields to capture local and global dependencies, and incorporates class weights into the cross-entropy loss during training to address class imbalance.