wavClusteR

wavClusteR identifies RNA–protein interaction sites from PAR-CLIP (Photoactivatable-Ribonucleoside-Enhanced Crosslinking and Immunoprecipitation) data across the transcriptome by distinguishing genuine PAR-CLIP-induced base transitions from sequencing errors and SNPs using statistical models.

Key Features:

• Transition-Centered Algorithm: Employs a sensitive transition-centered algorithm tailored for PAR-CLIP to detect base transitions indicative of crosslink sites and to separate them from other transition sources.
• Non-Parametric Mixture Model: Uses a non-parametric mixture model to discriminate genuine PAR-CLIP-induced transitions from noise such as sequencing errors and cell type-specific SNPs.
• Bayesian Framework: Implements a Bayesian network approach that associates posterior log-odds with observed transitions to quantify confidence in detected interactions.
• High-Resolution Binding Site Resolution: Resolves protein binding sites at high resolution and computes detailed cluster statistics for precise characterization of binding locations.
• False Discovery Rate Control: Integrates RNA-Seq data to compute conservative experimentally based false discovery rates, reducing false positives in identified sites.
• Comparative Performance: Demonstrates favorable performance relative to alternative algorithms for PAR-CLIP analysis.

Scientific Applications:

• Transcriptome-wide mapping: Enables transcriptome-wide identification of RNA–protein interaction sites from PAR-CLIP experiments.
• RNA-binding protein characterization: Supports characterization of protein binding preferences and cluster statistics for RNA-binding proteins.
• Experimental design and validation: Facilitates the design of subsequent validation experiments by providing high-resolution sites and confidence estimates.

Methodology:

Applies a non-parametric mixture model to separate genuine PAR-CLIP-induced transitions from sequencing errors and SNPs, followed by a Bayesian network framework that computes posterior log-odds and estimates cluster statistics; integrates RNA-Seq data to compute experimentally based false discovery rates.