iEnhancer-ECNN

iEnhancer-ECNN identifies and classifies DNA enhancers and predicts enhancer strength from sequence data using ensemble convolutional neural networks.

Key Features:

• Data Transformation Techniques: Uses one-hot encoding and k-mer representations to convert DNA sequences into model inputs.
• Model Construction: Implements ensembles of convolutional neural networks (ECNNs) to aggregate multiple CNN predictors.
• Two-layer Framework: Operates a two-layer framework with Layer 1 for enhancer identification (accuracy 0.769) and Layer 2 for enhancer strength classification (accuracy 0.678).
• Evaluation Metrics: Assesses performance using Area Under the Receiver Operating Characteristic Curve (AUC), sensitivity, and Matthews correlation coefficient (MCC).

Scientific Applications:

• Comparative Benchmarking: Outperforms existing state-of-the-art methods on the Liu et al. benchmark dataset with improvements in AUC, sensitivity, and MCC.
• Robustness and Stability: Reports increases in MCC of approximately 11.0%, 46.5%, and 65.0% respectively, with notable improvement in the enhancer classification layer.

Methodology:

Transforms sequences via one-hot encoding and k-mers and applies ensemble convolutional neural networks in a two-layer framework for identification (layer 1) and strength classification (layer 2).