regshape

Regshape: Transcription Factor Binding Site Prediction Integrating DNA Shape

Regshape integrates position-weight matrix (PWM) models with DNA shape-derived features to improve prediction of transcription factor (TF) binding sites by combining sequence-based nucleotide probabilities with local three-dimensional DNA structural information.

Key Features:

• PWM Integration: Combines traditional position-weight matrix models with additional structural parameters to refine TF binding specificity estimation.
• DNA Shape-Derived Features: Incorporates sequence-dependent DNA structural characteristics into a generalized regulatory score.
• Regulatory Potential Scoring: Computes composite scores integrating PWM and DNA shape features to enhance discrimination between true binding and non-binding sequences.
• Empirical Benchmarking: Evaluated across 75 vertebrate transcription factors, demonstrating improved prediction accuracy for 45% without significant accuracy loss for others.

Scientific Applications:

• Regulatory Genomics: Improves identification of transcription factor binding sites for analysis of TF–DNA interactions and gene regulation.

Methodology:

Regshape combines traditional PWM-based nucleotide position probabilities with DNA shape feature-based regulatory potential scores into a unified predictive framework. The integrated model was empirically assessed using datasets from 75 vertebrate transcription factors to compare predictive accuracy against PWM-only approaches.