RNA as drug target has gained momentum during the last twenty years, offering exciting possibilities in drug discovery. In 2022, we introduced HARIBOSS (harnessing ribonucleic acid–small molecule structures, https://hariboss.pasteur.cloud/), a database of RNA-small molecule structures, updated at a regular basis. Building on this first version, we now present HARIBOSS++, a major update that enhances the platform’s capabilities to better analyze RNA as a therapeutic target. First, we have expanded HARIBOSS++ database from RNA-only to more complex binding sites, including proteins and DNA. We have then introduced a novel analysis of RNA binding sites, i.e., the “RNA pocketome,” using a range of different techniques. We have also conducted further studies to characterize the different RNA families in the database. We classified interacting ligands using tools such as Molformer or molecular fingerprints to outline any specific pattern in RNA-SM interaction. Moreover, we have integrated available experimental binding affinity data and evaluated the accuracy of a widely used in silico method for binding affinity calculations, namely free energy perturbation. We have also benchmarked the accuracy of Boltz2, a recent deep-learning structure predictor approach, in its ability to predict the RNA 3D structure as well as the ligand pose in co-folded structures. Finally, we have been exploring the potential of RNA large language models embeddings to identify signals that link RNA conformations and their related function. HARIBOSS++ enriches available experimental data with modern tools from the computational community into a complete database for drug discovery.
Marengo et al. (Sun,) studied this question.