AI Designs Novel RNA Pseudoknots by Learning Constraints, Not Full 3D Structure
bravo_abad · x · 2026-08-28
A striking AI-for-science result on designing RNA pseudoknots—complex structures where base pairings cross and interlock.
The twist: the AI did not need to predict the final 3D shape atom by atom. Instead it searched for RNA sequences likely to form the right base-pair pattern—the structural "skeleton"—with a model trained on chemical measurements from 1 million RNAs judging whether candidates would fold correctly. Synthesized molecules, verified by cryo-EM, folded into well-defined, previously unseen 3D structures with extra interactions the AI never specified. Lesson: learning the right constraints can suffice to design a physical system.
Related event: Generative AI Designs Complex RNA Pseudoknots, Featured in Science(4 posts)→
More from Research
- Robotics papers on arXiv have grown ~6x in five years, peaking at 676 in a single week — Rewkang · 2026-10-15
- MicroSDF: microfacet-driven neural SDFs for mixed-reflectance 3D reconstruction — ssh4net · 2026-10-15
- 'Thinking leakage': no-think RL gains in hybrid models just reuse think-mode skills — kaust-generative-ai · 2026-10-15
- COMPASS predicts model-pair compatibility before alignment or merging — epfl-ch · 2026-10-15
- BERT tokenizer gets 50x speedup by skipping text normalization, not optimizing it — LysandreJik · 2026-10-15
- LLM agents autonomously design a SAR ADC on SKY130 from schematic to signoff, paper finds — IgorCarron · 2026-10-15