Tapioca: a platform for predicting de novo protein-protein interactions in dynamic contexts

Area of Science:

• Proteomics
• Systems Biology
• Bioinformatics

Background:

• Protein-protein interactions (PPIs) are crucial for cellular functions and responses.
• Understanding dynamic PPIs is essential for deciphering cellular states and disease mechanisms.

Purpose of the Study:

• To develop Tapioca, an ensemble machine learning framework for predicting global PPIs in dynamic cellular contexts.
• To improve experimental workflows for PPI studies, specifically thermal proximity coaggregation.
• To investigate viral infection dynamics by characterizing temporal PPIs.

Main Methods:

• Integration of mass spectrometry interactome data (thermal/ion denaturation, co-fractionation) with protein properties and functional networks.
• Optimization of thermal denaturation for increased throughput and cell lysis for enhanced protein detection.
• Application of the Tapioca workflow to study PPIs during Kaposi's sarcoma-associated herpesvirus reactivation.

Main Results:

• Tapioca successfully predicts de novo PPIs by integrating diverse data types.
• Experimental workflow improvements led to higher throughput and better detection of proteins from various subcellular compartments.
• Identification of NUCKS as a proviral hub protein during viral reactivation.
• Uncovered a broader role for NUCKS by integrating PPI networks across different herpesvirus families.

Conclusions:

• Tapioca offers a novel computational approach for studying PPIs in dynamic biological contexts.
• The optimized experimental workflow enhances the efficiency and scope of interactome studies.
• The findings provide insights into viral infection mechanisms and identify potential therapeutic targets.