iEnhancer-XG

iEnhancer-XG predicts enhancer regions and assesses their strength in genomic DNA to support studies of gene regulation.

Key Features:

• Two-Layer Predictive Model: A two-layer architecture first identifies enhancer regions and then evaluates the strength of identified enhancers.
• Base Classifier: Uses XGBoost as the classifier in both predictive layers.
• Feature Extraction Methods: Incorporates k-Spectrum Profile, Mismatch k-tuple, Subsequence Profile, Position-specific Scoring Matrix (PSSM), and Pseudo Dinucleotide Composition (PseDNC) to encode DNA sequences.
• Ensemble Learning: Integrates outputs from the five feature extraction methods into an ensemble framework to improve prediction accuracy.
• Interpretability with SHAP: Applies SHapley Additive exPlanations (SHAP) to provide feature-level interpretability for model predictions.
• Performance: Reported accuracies are 0.811 for enhancer identification (first layer) and 0.657 for strength prediction (second layer), evaluated using 10-fold cross-validation.
• Dependencies: Relies on repDNA for DNA sequence feature computation and SHAP for interpretability.

Scientific Applications:

• Enhancer Identification: Pinpoints enhancer regions within genomic sequences to facilitate studies of gene regulation.
• Strength Prediction: Assesses enhancer strength to infer potential regulatory impacts on gene expression.
• Comparative Analysis: Supports comparative evaluation against existing methods, with reported superior accuracy confirmed by 10-fold cross-validation.

Methodology:

Feature extraction using k-Spectrum Profile, Mismatch k-tuple, Subsequence Profile, PSSM, and PseDNC; integration of features via ensemble learning; classification with XGBoost in a two-layer architecture; interpretability via SHAP; performance assessed by 10-fold cross-validation.