henhaobio
Sign in

bioRxivBioinformaticsPreprint

Accurate prediction of nucleic acid and protein-nucleic acid complexes using RoseTTAFoldNA

Minkyung BaekiD, Ryan McHughiD, Ivan V. AnishchenkoiD, David A. BakeriD, Frank DiMaioiD

PDFDOI
Structured summary available · Jump to summary ↓

Abstract

Protein-nucleic acid complexes play critical roles in biology. Despite considerable recent advances in protein structure prediction, the prediction of the structures of protein-nucleic acid complexes without homology to known complexes is a largely unsolved problem. Here we extend the RoseTTAFold end-to-end deep learning approach to modeling of nucleic acid and protein-nucleic acid complexes. We develop a single trained network, RoseTTAFoldNA, that rapidly produces 3D structure models with confidence estimates for protein-DNA and protein-RNA complexes, and for RNA tertiary structures. In all three cases, confident predictions have considerably higher accuracy than current state of the art methods. RoseTTAFoldNA should be broadly useful for modeling the structure of naturally occurring protein-nucleic acid complexes, and for designing sequence specific RNA and DNA binding proteins.

Abstract summary

Claude Opus 5.5 · xhigh effort · 18 Oct 2026 · from the abstract

One RoseTTAFold-derived network models protein-DNA and protein-RNA complexes plus RNA tertiary structure end to end, returning confidence estimates alongside each 3D model.

Protein structure prediction has advanced quickly, but the abstract states that modelling protein-nucleic acid assemblies lacking homologues among solved complexes is still largely unsolved. Extending an end-to-end folding network to nucleic acids gives a single model that covers protein-DNA, protein-RNA and RNA-only cases.

Key results

  1. ClaimedFor all three prediction categories, the subset of models the network marks as confident is reported to be considerably more accurate than existing methods.

    Stated qualitatively in the abstract for protein-DNA complexes, protein-RNA complexes and RNA tertiary structures; no metric, benchmark set or named baseline method is given. · Abstract

  2. ClaimedA single trained network, RoseTTAFoldNA, handles nucleic acid and protein-nucleic acid modelling and emits a confidence estimate with each structure.

    Described as one network that rapidly produces 3D models with confidence estimates across the three categories; speed, architecture and training data are not quantified in the abstract. · Abstract

  3. ClaimedThe authors position the method both for modelling naturally occurring protein-nucleic acid complexes and for designing sequence-specific DNA- and RNA-binding proteins.

    Asserted in the closing sentence of the abstract; no design case study or experimental validation is reported in the available text. · Abstract

Methods, data and reproducibility

The RoseTTAFold end-to-end deep learning approach for protein structure prediction is extended to nucleic acids and trained as one network that outputs 3D coordinates with confidence estimates for protein-DNA complexes, protein-RNA complexes and RNA tertiary structures. The abstract gives no architecture details, training set or benchmark description.

Methods & tools
RoseTTAFoldNA, RoseTTAFold
Limitations (2)
  • Only a relative accuracy claim is available: the abstract names no metric, no evaluation set and no specific competing method, so the size of the improvement cannot be judged. · noted in this summary
  • The accuracy advantage is stated for confident predictions only, and the text does not say what fraction of inputs reach that confidence level or how the estimate is calibrated. · noted in this summary

Read this if you model protein-DNA or protein-RNA complexes or RNA tertiary structures and want to know what a nucleic-acid extension of RoseTTAFold claims to deliver.

Citations
60

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISMartinović '24Ludaic '25Kappel '18Roche '25Feng '22Wang '23
72% semantic matchBioinformatics
Ivona Martinović, Tin Vlašić, Yang Li+3 more·bioRxiv·Nov 27, 2024· 5 citations