RAPDOR: Using Jensen-Shannon Distance for the computational analysis of complex proteomics datasets

Abstract

Abstract The computational analysis of large proteomics datasets from gradient profiling or spatially resolved proteomics is often as crucial as experimental design. We present RAPDOR, a tool for intuitive analyzing and visualizing such datasets, based on the Jensen-Shannon distance and analysis of similarities between replicates, applied to the identification of RNA-binding proteins (RBPs) and spatial proteomics. First, we examine the in-gradient distribution profiles of protein complexes with or without RNase treatment (GradR) to identify RBPs in the cyanobacterium Synechocystis 6803. RBPs play pivotal regulatory and structural roles. Although numerous RBPs are well characterized, the complete set of RBPs remains unknown for any species. RAPDOR identifies 165 potential RBPs, including ribosomal proteins, RNA-modifying enzymes, and proteins not previously associated with RNA binding. High-ranking putative RBPs, such as ribosome hibernation factor LrtA/RaiA, phosphoglucomutase Sll0726, antitoxin Ssl2245, and preQ(1) synthase QueF predicted by RAPDOR but not the TriPepSVM algorithm, are experimentally validated, indicating the existence of uncharacterized RBP domains. These data are available online, providing a resource for RNase-sensitive protein complexes in cyanobacteria. We then show by reanalyzing existing datasets that RAPDOR effectively examines the intracellular redistribution of proteins upon growth factor stimulation. RAPDOR is a generic, non-parametric tool for analyzing highly complex datasets.