De novo design of RNA pseudoknots with deep learning

RNA design has been hindered by the limited accuracy of 3D structure prediction. Here, we show that intricate RNA structures can be generated with current deep learning tools through accurate de novo design of pseudoknot secondary structures. In an Eterna competition involving 57 pseudoknots, generative AI methods matched experienced human designers in solving most blind challenges, evaluated by single-nucleotide-resolution chemical mapping, compensatory mutagenesis, and cryogenic electron microscopy. AI-generated molecules with accurate secondary structures formed well-ordered 3D folds stabilized by noncanonical tertiary interactions not modeled during design. Success was guided by an RNet foundation model trained on prior chemical mapping data, suggesting that some difficult RNA design tasks may be tractable without first solving RNA 3D structure prediction.

Authors

Institutions

Publication Details

Journal
Science
Published
2026-08-27
DOI
https://doi.org/10.1126/science.aeg6829
Citations
2
Primary Topic
RNA and protein synthesis mechanisms
Type
article
Field-Weighted Citation Impact
6.06

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
article

De novo design of RNA pseudoknots with deep learning

Jigyasa Verma, Wipapat Kladwang, Jill Townley, Boris Rudolfs et al.
2 citations
Science
RNA and protein synthesis mechanisms
6.06
article

De novo design of RNA pseudoknots with deep learning

Jigyasa Verma, Wipapat Kladwang, Jill Townley, Boris Rudolfs, Po‐Ssu Huang, Thomas G. Karagianes, Jonathan Romano, Gina El Nesr, J. Hingey, Eli Fisker, Andrew Favor, Rhiju Das, Daniel B. Haack, Andrew Kubaney, Eterna Participants, Chaitanya K Joshi, R J Huang, Adamo Mancino, Pietro Lio, Christian Choe, Shujun He, David Baker, Vivian Wu, Nicholas Spellmon, Zhiheng Yu, Hamish M Blair, Navtej Toor
article en
2 citations

Abstract

RNA design has been hindered by the limited accuracy of 3D structure prediction. Here, we show that intricate RNA structures can be generated with current deep learning tools through accurate de novo design of pseudoknot secondary structures. In an Eterna competition involving 57 pseudoknots, generative AI methods matched experienced human designers in solving most blind challenges, evaluated by single-nucleotide-resolution chemical mapping, compensatory mutagenesis, and cryogenic electron microscopy. AI-generated molecules with accurate secondary structures formed well-ordered 3D folds stabilized by noncanonical tertiary interactions not modeled during design. Success was guided by an RNet foundation model trained on prior chemical mapping data, suggesting that some difficult RNA design tasks may be tractable without first solving RNA 3D structure prediction.

ScienceVol. 393(6814)
Howard Hughes Medical Institute (US), Janelia Research Campus (US), University of Washington (US), University of Cambridge (GB), University of California San Diego (US), CAE Solutions (United States) (US), Swift Engineering (United States) (US), Eterna Massive Open Laboratory (US), Texas A&M University (US), Stanford University (US)
National Science Foundation, Howard Hughes Medical Institute, National Institutes of Health
Openalex Percentile: Top 3%
RNA and protein synthesis mechanisms
6.06
DOIOpen Access PDFOpenAlex
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.