pyRBDome: a comprehensive computational platform for enhancing RNA-binding proteome data.

Area of Science:

• Molecular Biology
• Bioinformatics

Background:

• High-throughput proteomics methods have advanced RNA-binding protein (RBPome) and RNA-binding sequence (RBDome) identification.
• Quantifying noise, such as false positives, in these large datasets is challenging due to low-throughput validation methods.

Purpose of the Study:

• To introduce pyRBDome, an in silico pipeline designed to enhance the accuracy and reliability of RNA-binding proteome data.
• To improve the identification of genuine RNA-binding proteins and RNA-binding sites.

Main Methods:

• pyRBDome integrates experimental RBPome/RBDome data with RNA-binding site (RBS) predictions from machine-learning tools.
• It incorporates high-resolution structural data and employs statistical evaluation for data refinement.
• New ensemble machine-learning models were trained using pyRBDome results to boost RBS detection sensitivity and specificity.

Main Results:

• pyRBDome facilitates the rapid identification of likely true RNA-binders within experimental datasets.
• Analysis of human RBDome data revealed discrepancies between UV-cross-linked sites and actual RNA binding in high-resolution structures.
• The pipeline demonstrated enhanced sensitivity and specificity in RNA-binding site detection.

Conclusions:

• pyRBDome offers a robust computational approach to increase confidence in RNA-binding proteome and sequence datasets.
• It addresses limitations of using structural data as a sole benchmark for RNA-binding validation.
• The pipeline serves as a valuable tool for researchers working with RBPome and RBDome data.