csDMA

csDMA predicts DNA N6-methyldeoxyadenosine (6mA) modification sites genome-wide using machine learning to identify epigenetic marks.

Key Features:

• Feature Encoding Schemes: Generates a feature matrix using Motif, Kmer, and Binary encoding schemes to transform DNA sequences for machine learning.
• Machine Learning Algorithms: Evaluates multiple algorithms and identifies the ExtraTrees algorithm as particularly effective for 6mA prediction.
• Performance Metrics: Reports AUC values of 0.878 from 5-fold cross-validation on the training dataset and 0.893 on an independent testing dataset.
• Benchmarking: Compares predictive performance against existing state-of-the-art tools and reports superior performance.
• Implementation: Implemented in Python 2.7.

Scientific Applications:

• Large-scale epigenomic surveys: Enables computational identification of 6mA sites for genome-wide methylation mapping.
• Gene regulation studies: Facilitates investigation of 6mA roles in transcriptional regulation.
• Developmental biology research: Supports studies of 6mA dynamics during development.
• Disease mechanism investigations: Assists exploration of 6mA-associated mechanisms in disease across various organisms.

Methodology:

Constructs a feature matrix using Motif, Kmer, and Binary encodings; evaluates multiple machine learning algorithms and selects ExtraTrees; assesses performance via 5-fold cross-validation (AUC 0.878) and independent testing (AUC 0.893); benchmarks against state-of-the-art tools.