Predicting RNA 3D structure and conformers using a pre-trained secondary structure model and structure-aware attention

2026-04-21 · Source: Machine learning : nature.com subject feeds · Field: Science & Research — Life Sciences & Biology, Mathematics & Computational Sciences, Artificial Intelligence & Machine Learning · Depth: Expert, long

Summary

trRosettaRNA2, a deep learning-based end-to-end approach, addresses the challenge of predicting RNA three-dimensional (3D) structure and conformers, which is difficult due to scarce experimental data and RNA flexibility. Published on April 21, 2026, in Nature Machine Intelligence, this method integrates an auxiliary secondary structure (SS) prior module, pre-trained on extensive SS data, to generate informative base-pairing priors. This module also functions as an independent SS prediction tool, trRNA2-SS, achieving state-of-the-art performance. trRosettaRNA2 employs SS-aware attention to predict RNA 3D structures and conformers, outperforming other methods in rigorous benchmarks while using fewer parameters and computational resources. The Yang-Server group, utilizing trRosettaRNA2, was the top automated server for RNA structure prediction in the CASP16 blind test, surpassing AlphaFold 3. Its application to ribonuclease P RNA successfully captures structural heterogeneity without experimental data, demonstrating its potential for predicting RNA conformational ensembles.

Key takeaway

For AI Scientists and Research Scientists working on structural biology, trRosettaRNA2 offers a robust solution for RNA 3D structure and conformer prediction. Its superior performance in CASP16, even against AlphaFold 3, suggests it can significantly enhance your ability to model complex RNA molecules and their dynamic behaviors. You should consider integrating this tool into your research workflows, especially when experimental data is limited, to explore RNA conformational ensembles more effectively.

Key insights

trRosettaRNA2 accurately predicts RNA 3D structures and conformers by integrating pre-trained secondary structure priors and SS-aware attention.

Principles

• Leverage secondary structure priors for 3D prediction.
• Integrate SS-aware attention for end-to-end modeling.
• Predict conformational ensembles for intrinsic flexibility.

Method

trRosettaRNA2 uses a deep learning pipeline with an auxiliary secondary structure (SS) prior module, pre-trained on extensive SS data, and SS-aware attention to generate RNA 3D structures and conformers.

In practice

• Use trRosettaRNA2 for RNA 3D structure prediction.
• Apply trRNA2-SS for RNA secondary structure prediction.
• Explore RNA conformational landscapes with diverse SS inputs.

Topics

• RNA 3D Structure Prediction
• RNA Conformational Ensembles
• Deep Learning
• trRosettaRNA2
• Secondary Structure Prior

Code references

• YangLab-SDU/trRosettaRNA2

Best for: AI Scientist, Research Scientist

Related on AIssential

• RNA design across eras: from covariance models to modern generative AI — RNA's functional versatility relies on 3D folding, with generative AI building on foundational probabilistic models for structure prediction.
• Two-dimensional geometric template diffusion for boosting single-sequence protein structure prediction — TDFold uses 2D geometric template diffusion for efficient, accurate single-sequence protein structure prediction.
• Bridging three-dimensional molecular structures and artificial intelligence with a conformation description language — ConfSeq enables large language models to process 3D molecular structures by converting them into discrete token sequences.
• Polyformer: a generative framework for thermodynamic modeling of polymeric molecules — Polyformer generates thermodynamic conformational ensembles for polymeric molecules, predicting folding and temperature-dependent changes.
• Unraveling the Mechanism of Drug Binding to SARS-CoV-2 RNA Pseudoknot with Thermodynamics-Driven Machine Learning — Thermodynamics-driven machine learning reveals topology-selective drug binding mechanisms to SARS-CoV-2 RNA pseudoknot.
• From possibility to precision in macromolecular ensemble prediction — Predicting dynamic protein conformational ensembles requires integrating diverse experimental data and advanced AI.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.