RNA-binding protein transcripts reflect composition of target mRNAs.

RNA-protein interactions are central to gene regulation, yet the large-scale organization of RNA-protein networks remains incompletely understood. Using a comprehensive human eCLIP dataset encompassing interactions between 150 RNA-binding proteins (RBPs) and >11 000 mRNAs, we identify a robust organizing principle underlying the RNA-protein network structure: mRNAs preferentially associate with RBPs whose own encoding transcripts share similar nucleotide composition. In other words, mRNAs enriched in a given nucleotide tend to be targeted by RBPs whose own transcripts are likewise enriched in that nucleotide and vice versa. This global pattern holds for all four RNA nucleotides, remains statistically significant after controlling for transcript length, expression levels and sequence-motif-driven interactions, and is confirmed by in vitro HTR-SELEX data. We use the observed relationship to rationalize the spatial organization of mRNAs in the nucleus i.e. the known G/C gradient towards nuclear speckles. Notably, an mRNA's propensity toward RBPs rich in arginine, which is predominantly encoded by and preferentially binds guanine, is a strong predictor of its speckle enrichment. Our findings highlight a fundamental link between coding and binding in biology and suggest that mRNA composition biases provide a fundamental layer of specificity in shaping the global RNA-protein interaction network.