Unraveling biomolecular and community grammars of RNA granules via machine learning

Membraneless RNA granules are essential for posttranscriptional gene regulation, influencing cellular functions and contributing to neurodegenerative diseases. However, a comprehensive understanding of their compositions and organization has been challenging due to their complex nature. In this study, we develop robust machine learning models to reliably identify RNA granule proteomes within the human proteome, capturing central RNA granule characteristics despite the heterogeneity across diverse in vitro conditions. Furthermore, we uncover protein-protein interaction (PPI) community grammars within the RNA granule proteome, highlighting PPIs as key stabilizers of RNA granule structure and function. Dense PPI clusters serve as stable "cores," forming key functional subunits across heterogeneous RNA granules. We introduce a state-of-the-art framework for understanding RNA granule biology and underscore the critical role of PPIs in maintaining RNA granule integrity.