黑料网爆吃瓜

Join D2R for this hybrid seminar with Professor Carlos Oliver from Vanderbilt University about "Structure-Based Machine Learning for RNA Drug Discovery: The Story So Far".聽

RNA has emerged as a vast and underexplored class of therapeutic targets, yet RNA-targeted drug discovery has lagged behind its protein counterpart. The reasons are well-rehearsed: RNA structural data is scarce, RNA-ligand interactions are governed by distinct biophysics, and the machine learning toolkit that powered the protein-folding revolution does not transfer cleanly to RNA. Specific small-molecule recognition of RNA is achievable, but unlocking the potential of machine learning for RNA drug discovery will require specialized methods at every step of the workflow.

In this talk, Prof Oliver will describe such a pipeline. In target discovery, substructure-aware methods learn the recurring structural building blocks of RNA and the binding pockets where small molecules engage: graph representation learning recovers and expands known motif libraries, and adaptive graph coarsening predicts druggable binding sites that already support experimental virtual screens. In hit identification, structure-based virtual screening on coarse-grained, base-pair-aware representations runs orders of magnitude faster than docking and has produced the first experimentally validated hits from a structure-based deep learning system. In evaluation, shared benchmarking infrastructure for RNA structure-function modeling, alongside a recent field review proposing standards for measuring drug-target specificity, gives the community common ground to compare approaches. The recurring theme is that respecting the modality, through coarse-grained representations, RNA-specific self-supervision, and substructure-aware reasoning, is what unlocks data efficiency and wet-lab translation.

Agenda

Register

For those attending in-person, confirmation emails will be provided to registered participants a few days before the event.聽Space is limited!聽