A comparative analysis of RNA-binding proteins binding models learned from RNAcompete, RNA Bind-n-Seq and eCLIP data

Area of Science:

• Molecular Biology
• Genomics
• Bioinformatics

Background:

• Post-transcriptional gene regulation is crucial in biology.
• RNA-binding proteins (RBPs) play a significant role in this process.
• High-throughput technologies like RNAcompete and RNA Bind-n-Seq measure RBP binding to RNA sequences.

Purpose of the Study:

• To compare the RNAcompete and RNA Bind-n-Seq technologies for measuring RBP binding.
• To assess the concordance between binding models derived from these two methods.
• To evaluate the predictive power of these models for in vivo RBP binding and RNA structural preferences.

Main Methods:

• Inferred $k$-mer binding models from RNAcompete and RNA Bind-n-Seq data for 31 human RBPs.
• Calculated Pearson correlation to assess model agreement and predictive performance.
• Measured the area under the curve (AUC) to evaluate in vivo binding prediction accuracy.
• Analyzed RNA structural preferences inferred from binding models.

Main Results:

• RNA Bind-n-Seq and RNAcompete models showed agreement (Pearson correlation > 0.5) for 23 out of 31 RBPs.
• RNA Bind-n-Seq models predicted RNAcompete binding well (average Pearson correlation 0.26).
• Both technologies achieved comparable performance in predicting in vivo binding (average AUC 0.7) and showed high concordance in RNA structural preferences.

Conclusions:

• RNAcompete and RNA Bind-n-Seq are largely consistent in modeling RBP binding and predicting in vivo interactions.
• The study developed a new $k$-mer score for RNA Bind-n-Seq, incorporating RNA structural preferences.
• These findings validate and enhance the utility of these technologies for studying post-transcriptional gene regulation.