Live-Cell Covalent Profiling Reveals Principles of RNA-Small Molecule Recognition across the Human Transcriptome

RNA folds are abundant in mammalian cells yet poorly characterized as small-molecule targets. We present a scalable, unbiased live-cell pipeline that maps where small molecules bind RNA across the human transcriptome and convert those binders into selective degraders. A 200-member fragment library bearing diazirine/alkyne handles yielded 23 RNA-binding candidates. Chem-CLIP-Map-Seq in MDA-MB-231 cells identified 723 RNA targets and their binding sites, revealing a strong bias toward 5′ and 3′ untranslated regions (UTRs) in mRNAs and enrichment at thermodynamically stable structures, with limited binding to non-coding RNAs. Expression level and local stability contributed to engagement. An integrated machine-learning model trained on multiple fingerprints distinguished binders from non-binders, and highlighted chemotypes and physicochemical features that favor RNA recognition. Four fragments were converted to RiboTACs; despite broad binding, cleavage was highly selective, with X1-RiboTAC degrading MPP7 and SSC4D mRNAs in an RNase L-dependent manner and reducing their protein levels. A competitive profiling workflow quantified in-cell target occupancy and guided optimization of the RNA-binding module to reprogram selectivity: an X1 derivative produced an MPP7 -selective RiboTAC that lowered MPP7 mRNA levels and suppressed breast-cancer cell migration, while sparing SSC4D transcripts. This end-to-end framework, including transcriptome-wide mapping, data-driven rules, and tunable degradation, establishes practical principles for ligandable RNA sites in cells and enables rational design of RNA-targeted small molecules and degraders.