RNA-binding proteins (RBPs) are the cornerstone for nearly all RNA cellular functions from transcription to forming integral parts of sophisticated structures such as spliceosomes and ribosomes. Therefore, RNA-protein interactions are diverse and often complex. In some cases, the interaction of the RBP binding sites and the RNA can be transient and highly dynamic with the RNA adopting multiple folded structures. As a result, RBPs are often viewed as challenging targets for drug development, despite their significant roles in various biological processes such as transcription, splicing, translation, and protein localization. A deep understanding of these interactions and identification of patterns associated the RNA-protein interface may open opportunities to target RBPs for a wide array of therapeutic and biotechnological applications.
Inhibitors against a handful of RBPs have been reported (FtsY, Hfq, Cas9, CsrA, RNase P, among others) but it remains to be seen whether general strategies for targeting RBPs can be developed. In this study, we analyzed the structural and physicochemical properties of the interface regions in experimentally solved bacterial RNA-protein structures. Using a combination of machine learning and deep learning, we clustered these complexes based on their interaction characteristics. This AI-based approach grouped two of our previously selected targets (FtsY, Hfq) based on visual inspection together with other potential RBP targets, demonstrating its efficacy. For the newly selected targets, molecular dynamics (MD) simulations will be employed to model the binding events at the atomic level. Our findings so far suggest that a significant portion of bacterial RBPs can potentially be targeted through small molecules, peptides, or modified RNAs. This hybrid approach provides a framework for identifying and investigating RBP interactions, paving the way for novel therapeutic strategies.