i6mA-VC

i6mA-VC predicts DNA N6-methyladenine (6 mA) sites using machine learning to identify epigenetic modifications relevant to gene expression regulation.

Key Features:

• Multi-Classifier Voting Methodology: Uses a voting ensemble that integrates Gradient Boosting Decision Tree (GBDT), Light Gradient Boosting Machine (LightGBM), and Multilayer Perceptron Classifier (MLPC) to improve prediction robustness.
• k-mer and Binary Encoding with GBDT Selection: Extracts sequence-based features via k-mer and binary encoding and refines them using a GBDT-embedded feature selection method.
• Ring-Function-Hydrogen-Chemical Properties (RFHCP) with ExtraTree Selection: Encodes DNA sequences as RFHCP vectors and applies an ExtraTree strategy for feature selection.
• Cross-Species Prediction Capability: Supports applying models across species to address dataset-specific limitations and improve generalization.
• Validation Strategy: Evaluates model performance using five-fold cross-validation.

Scientific Applications:

• Performance on specific species: Achieves reported accuracies of 0.888 for the Rice dataset, 0.967 for the M.musculus dataset, and 0.848 for cross-species datasets.
• Epigenetic and gene regulation studies: Facilitates identification of 6 mA sites to support research into DNA modification processes and gene expression regulation.

Methodology:

Combines k-mer/binary encoding and RFHCP feature representations with feature selection via a GBDT-embedded method and an ExtraTree strategy, fuses selected features into an ensemble of GBDT, LightGBM, and MLPC using a voting scheme, and validates performance via five-fold cross-validation.