Enhancer-FRL

Enhancer-FRL implements a two-layer prediction framework to distinguish enhancers from non-enhancers and predict enhancer activity strength (strong or weak) for studies of gene regulation.

Key Features:

• Two-Layer Prediction Model: Distinguishes enhancers versus non-enhancers in the first layer and classifies enhancer activity strength (strong or weak) in the second layer.
• Feature Representation Learning Scheme: Encodes ten distinct features and integrates them with five machine learning algorithms to produce a 50-dimensional (50D) probabilistic vector.
• Multiview Probabilistic Integration: Integrates multiview probabilistic features to construct the final prediction model, improving robustness relative to single-feature representations.
• Performance Improvement: Demonstrates significant performance gains over existing state-of-the-art tools as evidenced by assessments on independent test datasets.

Scientific Applications:

• Enhancer identification and activity prediction: Enables detection of genomic enhancers and prediction of their activity strength for downstream regulatory analyses.
• Gene regulation studies: Supports investigation of regulatory elements and their roles in controlling gene expression.
• Large-scale genomic data analysis: Facilitates precise enhancer characterization in genome-wide datasets.

Methodology:

Ten distinct features are encoded; five machine learning algorithms generate a 50D probabilistic vector from those features; multiview probabilistic features are integrated to form the final prediction model; a two-layer predictor separates enhancers from non-enhancers and classifies enhancer strength.