rewirebio.iobenchmarks
Configuration

RhoFold+

RhoFold+ predicts RNA three-dimensional structures from RNA sequence with alignment information.

Sources (3)ml4bio/RhoFold: README.md; cuhkaih/rhofold: README.md; rhofold: Journal full-text XML · Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata

6 evaluations · 12 results

How it worksRhoFold+ workflow
RhoFold+ workflow1. RNA sequence and MSA. Then: 2. RhoFold+ prediction. Then: 3. Structure and confidence. Then: 4. Optional relaxationRhoFold+ workflow1. RNA sequence and MSA. Then: 2. RhoFold+ prediction. Then: 3. Structure and confidence. Then: 4. Optional relaxationRhoFold+ workflow1. RNA sequence and MSA. Then: 2. RhoFold+ prediction. Then: 3. Structure and confidence. Then: 4. Optional relaxation

Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.

Sources (3)ml4bio/RhoFold: README.md; cuhkaih/rhofold: README.md; rhofold: Journal full-text XML · Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata

Overview

Model type

RNA structure predictor with language-model and MSA inputs

Inputs

RNA FASTA (A, U, G, C), optionally a supplied MSA; an MSA is otherwise generated by the full workflow.

Outputs

Three-dimensional PDB models, predicted distograms, secondary structure and per-residue confidence.

Sources (3)ml4bio/RhoFold: README.md; cuhkaih/rhofold: README.md; rhofold: Journal full-text XML · Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata

limited source coverage · Automated source review, 2026-09-16. All specifications and missing details

Model Capabilities & Hardware Requirements

Deployment facts recorded for RhoFold+. Fields the record does not carry are marked not recorded.

ParametersNot recorded
Minimum VRAMNot recorded
Inference LatencyNot recorded
Context WindowNot recorded
AccessNot recorded
LicenseNot recorded

Evaluations and results

6 evaluations · 12 results. Different protocols are not a single leaderboard.

Sorted by Casp sum z score (higher is better). The best value in each column is highlighted. Decimals are rounded for display; each value links to the printed value and its source.

Tested configurationProtocol and datasetCasp sum z scoreRMSD
RhoFold+EvaluationRhoFold+ 2024 CASP15 natural RNA comparison R1107-sum_z_gdt_ts_tm_score: R1107: Sum of Z-scores for GDT-TS and TM-score, CASP15 natural RNA target R1107 (RhoFold+ 2024 CASP15 natural RNA comparison split)4.68 (best in column)Not reported
RhoFold+EvaluationRhoFold+ 2024 CASP15 natural RNA comparison R1108-sum_z_gdt_ts_tm_score: R1108: Sum of Z-scores for GDT-TS and TM-score, CASP15 natural RNA target R1108 (RhoFold+ 2024 CASP15 natural RNA comparison split)4.52Not reported
RhoFold+EvaluationRhoFold+ 2024 CASP15 natural RNA comparison R1149-sum_z_gdt_ts_tm_score: R1149: Sum of Z-scores for GDT-TS and TM-score, CASP15 natural RNA target R1149 (RhoFold+ 2024 CASP15 natural RNA comparison split)3.75Not reported
RhoFold+EvaluationRhoFold+ 2024 CASP15 natural RNA comparison R1116-sum_z_gdt_ts_tm_score: R1116: Sum of Z-scores for GDT-TS and TM-score, CASP15 natural RNA target R1116 (RhoFold+ 2024 CASP15 natural RNA comparison split)2.68Not reported
RhoFold+EvaluationRhoFold+ 2024 CASP15 natural RNA comparison R1117-sum_z_gdt_ts_tm_score: R1117: Sum of Z-scores for GDT-TS and TM-score, CASP15 natural RNA target R1117 (RhoFold+ 2024 CASP15 natural RNA comparison split)2.01Not reported
RhoFold+EvaluationRhoFold+ 2024 CASP15 natural RNA comparison R1156-sum_z_gdt_ts_tm_score: R1156: Sum of Z-scores for GDT-TS and TM-score, CASP15 natural RNA target R1156 (RhoFold+ 2024 CASP15 natural RNA comparison split)1.32Not reported
RhoFold+EvaluationRhoFold+ 2024 CASP15 natural RNA comparison R1107-rmsd: R1107: RMSD, CASP15 natural RNA target R1107 (RhoFold+ 2024 CASP15 natural RNA comparison split)Not reported5.01 Å
RhoFold+EvaluationRhoFold+ 2024 CASP15 natural RNA comparison R1108-rmsd: R1108: RMSD, CASP15 natural RNA target R1108 (RhoFold+ 2024 CASP15 natural RNA comparison split)Not reported4.54 Å
RhoFold+EvaluationRhoFold+ 2024 CASP15 natural RNA comparison R1116-rmsd: R1116: RMSD, CASP15 natural RNA target R1116 (RhoFold+ 2024 CASP15 natural RNA comparison split)Not reported8.92 Å
RhoFold+EvaluationRhoFold+ 2024 CASP15 natural RNA comparison R1117-rmsd: R1117: RMSD, CASP15 natural RNA target R1117 (RhoFold+ 2024 CASP15 natural RNA comparison split)Not reported2.87 Å (best in column)
RhoFold+EvaluationRhoFold+ 2024 CASP15 natural RNA comparison R1149-rmsd: R1149: RMSD, CASP15 natural RNA target R1149 (RhoFold+ 2024 CASP15 natural RNA comparison split)Not reported7.66 Å
RhoFold+EvaluationRhoFold+ 2024 CASP15 natural RNA comparison R1156-rmsd: R1156: RMSD, CASP15 natural RNA target R1156 (RhoFold+ 2024 CASP15 natural RNA comparison split)Not reported17.43 Å
All 12 result rows with coverage, uncertainty and sources
Filter evaluations

Applied filters: All linked evaluations

Exact evaluated configurations and original reported results
Tested configurationProtocol and datasetFindingEvidence and details
Configuration: RhoFold+Protocol: RhoFold+ 2024 CASP15 natural RNA comparison R1107-rmsd: R1107: RMSD
Dataset subset: CASP15 natural RNA target R1107 (RhoFold+ 2024 CASP15 natural RNA comparison split)
5.01 Å RMSD
angstrom · lower

Uncertainty: Not yet extracted

Coverage: 1/1 CASP15 natural RNA target scored (Per-cell target count; source retains up to five candidate structures, not five independent evaluation targets.)

Author-reported evaluation · Source checked
Methods, coverage and source

RhoFold+ on RhoFold+ 2024 CASP15 natural RNA comparison R1107-rmsd: R1107: RMSD

Retrospective evaluation of six CASP15 natural single-chain RNA targets; each task retains one named target. Artificially designed targets and RNA-protein complexes excluded. RNA sequence inputs without ions or other molecules. Paper Par13 selects the highest-performing of five candidates; Par51 describes five sampled-MSA runs for RhoFold+, one AlphaFold3 server run yielding five models, and five CASP15 models for the other methods. Figure 2h reports RMSD and sum of Z-scores for GDT-TS and TM-score. Metric-specific winning candidate identity, exact scoring-tool revision and RMSD atom selection remain unextracted. No target pooling or score recomputation.

Aggregation: Source-reported selected candidate for one target; no cross-target aggregation.

RhoFold+ Nature Methods 2024: Source Data Figure 2h · fig2hi!B11; method fig2hi!A11; target fig2hi!B1; Fig. 2h caption; XML Par13–15 and Par51
Configuration: RhoFold+Protocol: RhoFold+ 2024 CASP15 natural RNA comparison R1107-sum_z_gdt_ts_tm_score: R1107: Sum of Z-scores for GDT-TS and TM-score
Dataset subset: CASP15 natural RNA target R1107 (RhoFold+ 2024 CASP15 natural RNA comparison split)
4.68 Casp sum z score
unitless · higher

Uncertainty: Not yet extracted

Coverage: 1/1 CASP15 natural RNA target scored (Per-cell target count; source retains up to five candidate structures, not five independent evaluation targets.)

Author-reported evaluation · Source checked
Methods, coverage and source

RhoFold+ on RhoFold+ 2024 CASP15 natural RNA comparison R1107-sum_z_gdt_ts_tm_score: R1107: Sum of Z-scores for GDT-TS and TM-score

Retrospective evaluation of six CASP15 natural single-chain RNA targets; each task retains one named target. Artificially designed targets and RNA-protein complexes excluded. RNA sequence inputs without ions or other molecules. Paper Par13 selects the highest-performing of five candidates; Par51 describes five sampled-MSA runs for RhoFold+, one AlphaFold3 server run yielding five models, and five CASP15 models for the other methods. Figure 2h reports RMSD and sum of Z-scores for GDT-TS and TM-score. Metric-specific winning candidate identity, exact scoring-tool revision and RMSD atom selection remain unextracted. No target pooling or score recomputation.

Aggregation: Source-reported selected candidate for one target; no cross-target aggregation.

RhoFold+ Nature Methods 2024: Source Data Figure 2h · fig2hi!H11; method fig2hi!A11; target fig2hi!H1; Fig. 2h caption; XML Par13–15 and Par51
Configuration: RhoFold+Protocol: RhoFold+ 2024 CASP15 natural RNA comparison R1108-rmsd: R1108: RMSD
Dataset subset: CASP15 natural RNA target R1108 (RhoFold+ 2024 CASP15 natural RNA comparison split)
4.54 Å RMSD
angstrom · lower

Uncertainty: Not yet extracted

Coverage: 1/1 CASP15 natural RNA target scored (Per-cell target count; source retains up to five candidate structures, not five independent evaluation targets.)

Author-reported evaluation · Source checked
Methods, coverage and source

RhoFold+ on RhoFold+ 2024 CASP15 natural RNA comparison R1108-rmsd: R1108: RMSD

Retrospective evaluation of six CASP15 natural single-chain RNA targets; each task retains one named target. Artificially designed targets and RNA-protein complexes excluded. RNA sequence inputs without ions or other molecules. Paper Par13 selects the highest-performing of five candidates; Par51 describes five sampled-MSA runs for RhoFold+, one AlphaFold3 server run yielding five models, and five CASP15 models for the other methods. Figure 2h reports RMSD and sum of Z-scores for GDT-TS and TM-score. Metric-specific winning candidate identity, exact scoring-tool revision and RMSD atom selection remain unextracted. No target pooling or score recomputation.

Aggregation: Source-reported selected candidate for one target; no cross-target aggregation.

RhoFold+ Nature Methods 2024: Source Data Figure 2h · fig2hi!C11; method fig2hi!A11; target fig2hi!C1; Fig. 2h caption; XML Par13–15 and Par51
Configuration: RhoFold+Protocol: RhoFold+ 2024 CASP15 natural RNA comparison R1108-sum_z_gdt_ts_tm_score: R1108: Sum of Z-scores for GDT-TS and TM-score
Dataset subset: CASP15 natural RNA target R1108 (RhoFold+ 2024 CASP15 natural RNA comparison split)
4.52 Casp sum z score
unitless · higher

Uncertainty: Not yet extracted

Coverage: 1/1 CASP15 natural RNA target scored (Per-cell target count; source retains up to five candidate structures, not five independent evaluation targets.)

Author-reported evaluation · Source checked
Methods, coverage and source

RhoFold+ on RhoFold+ 2024 CASP15 natural RNA comparison R1108-sum_z_gdt_ts_tm_score: R1108: Sum of Z-scores for GDT-TS and TM-score

Retrospective evaluation of six CASP15 natural single-chain RNA targets; each task retains one named target. Artificially designed targets and RNA-protein complexes excluded. RNA sequence inputs without ions or other molecules. Paper Par13 selects the highest-performing of five candidates; Par51 describes five sampled-MSA runs for RhoFold+, one AlphaFold3 server run yielding five models, and five CASP15 models for the other methods. Figure 2h reports RMSD and sum of Z-scores for GDT-TS and TM-score. Metric-specific winning candidate identity, exact scoring-tool revision and RMSD atom selection remain unextracted. No target pooling or score recomputation.

Aggregation: Source-reported selected candidate for one target; no cross-target aggregation.

RhoFold+ Nature Methods 2024: Source Data Figure 2h · fig2hi!I11; method fig2hi!A11; target fig2hi!I1; Fig. 2h caption; XML Par13–15 and Par51
Configuration: RhoFold+Protocol: RhoFold+ 2024 CASP15 natural RNA comparison R1116-rmsd: R1116: RMSD
Dataset subset: CASP15 natural RNA target R1116 (RhoFold+ 2024 CASP15 natural RNA comparison split)
8.92 Å RMSD
angstrom · lower

Uncertainty: Not yet extracted

Coverage: 1/1 CASP15 natural RNA target scored (Per-cell target count; source retains up to five candidate structures, not five independent evaluation targets.)

Author-reported evaluation · Source checked
Methods, coverage and source

RhoFold+ on RhoFold+ 2024 CASP15 natural RNA comparison R1116-rmsd: R1116: RMSD

Retrospective evaluation of six CASP15 natural single-chain RNA targets; each task retains one named target. Artificially designed targets and RNA-protein complexes excluded. RNA sequence inputs without ions or other molecules. Paper Par13 selects the highest-performing of five candidates; Par51 describes five sampled-MSA runs for RhoFold+, one AlphaFold3 server run yielding five models, and five CASP15 models for the other methods. Figure 2h reports RMSD and sum of Z-scores for GDT-TS and TM-score. Metric-specific winning candidate identity, exact scoring-tool revision and RMSD atom selection remain unextracted. No target pooling or score recomputation.

Aggregation: Source-reported selected candidate for one target; no cross-target aggregation.

RhoFold+ Nature Methods 2024: Source Data Figure 2h · fig2hi!D11; method fig2hi!A11; target fig2hi!D1; Fig. 2h caption; XML Par13–15 and Par51
Configuration: RhoFold+Protocol: RhoFold+ 2024 CASP15 natural RNA comparison R1116-sum_z_gdt_ts_tm_score: R1116: Sum of Z-scores for GDT-TS and TM-score
Dataset subset: CASP15 natural RNA target R1116 (RhoFold+ 2024 CASP15 natural RNA comparison split)
2.68 Casp sum z score
unitless · higher

Uncertainty: Not yet extracted

Coverage: 1/1 CASP15 natural RNA target scored (Per-cell target count; source retains up to five candidate structures, not five independent evaluation targets.)

Author-reported evaluation · Source checked
Methods, coverage and source

RhoFold+ on RhoFold+ 2024 CASP15 natural RNA comparison R1116-sum_z_gdt_ts_tm_score: R1116: Sum of Z-scores for GDT-TS and TM-score

Retrospective evaluation of six CASP15 natural single-chain RNA targets; each task retains one named target. Artificially designed targets and RNA-protein complexes excluded. RNA sequence inputs without ions or other molecules. Paper Par13 selects the highest-performing of five candidates; Par51 describes five sampled-MSA runs for RhoFold+, one AlphaFold3 server run yielding five models, and five CASP15 models for the other methods. Figure 2h reports RMSD and sum of Z-scores for GDT-TS and TM-score. Metric-specific winning candidate identity, exact scoring-tool revision and RMSD atom selection remain unextracted. No target pooling or score recomputation.

Aggregation: Source-reported selected candidate for one target; no cross-target aggregation.

RhoFold+ Nature Methods 2024: Source Data Figure 2h · fig2hi!J11; method fig2hi!A11; target fig2hi!J1; Fig. 2h caption; XML Par13–15 and Par51
Configuration: RhoFold+Protocol: RhoFold+ 2024 CASP15 natural RNA comparison R1117-rmsd: R1117: RMSD
Dataset subset: CASP15 natural RNA target R1117 (RhoFold+ 2024 CASP15 natural RNA comparison split)
2.87 Å RMSD
angstrom · lower

Uncertainty: Not yet extracted

Coverage: 1/1 CASP15 natural RNA target scored (Per-cell target count; source retains up to five candidate structures, not five independent evaluation targets.)

Author-reported evaluation · Source checked
Methods, coverage and source

RhoFold+ on RhoFold+ 2024 CASP15 natural RNA comparison R1117-rmsd: R1117: RMSD

Retrospective evaluation of six CASP15 natural single-chain RNA targets; each task retains one named target. Artificially designed targets and RNA-protein complexes excluded. RNA sequence inputs without ions or other molecules. Paper Par13 selects the highest-performing of five candidates; Par51 describes five sampled-MSA runs for RhoFold+, one AlphaFold3 server run yielding five models, and five CASP15 models for the other methods. Figure 2h reports RMSD and sum of Z-scores for GDT-TS and TM-score. Metric-specific winning candidate identity, exact scoring-tool revision and RMSD atom selection remain unextracted. No target pooling or score recomputation.

Aggregation: Source-reported selected candidate for one target; no cross-target aggregation.

RhoFold+ Nature Methods 2024: Source Data Figure 2h · fig2hi!E11; method fig2hi!A11; target fig2hi!E1; Fig. 2h caption; XML Par13–15 and Par51
Configuration: RhoFold+Protocol: RhoFold+ 2024 CASP15 natural RNA comparison R1117-sum_z_gdt_ts_tm_score: R1117: Sum of Z-scores for GDT-TS and TM-score
Dataset subset: CASP15 natural RNA target R1117 (RhoFold+ 2024 CASP15 natural RNA comparison split)
2.01 Casp sum z score
unitless · higher

Uncertainty: Not yet extracted

Coverage: 1/1 CASP15 natural RNA target scored (Per-cell target count; source retains up to five candidate structures, not five independent evaluation targets.)

Author-reported evaluation · Source checked
Methods, coverage and source

RhoFold+ on RhoFold+ 2024 CASP15 natural RNA comparison R1117-sum_z_gdt_ts_tm_score: R1117: Sum of Z-scores for GDT-TS and TM-score

Retrospective evaluation of six CASP15 natural single-chain RNA targets; each task retains one named target. Artificially designed targets and RNA-protein complexes excluded. RNA sequence inputs without ions or other molecules. Paper Par13 selects the highest-performing of five candidates; Par51 describes five sampled-MSA runs for RhoFold+, one AlphaFold3 server run yielding five models, and five CASP15 models for the other methods. Figure 2h reports RMSD and sum of Z-scores for GDT-TS and TM-score. Metric-specific winning candidate identity, exact scoring-tool revision and RMSD atom selection remain unextracted. No target pooling or score recomputation.

Aggregation: Source-reported selected candidate for one target; no cross-target aggregation.

RhoFold+ Nature Methods 2024: Source Data Figure 2h · fig2hi!K11; method fig2hi!A11; target fig2hi!K1; Fig. 2h caption; XML Par13–15 and Par51
Configuration: RhoFold+Protocol: RhoFold+ 2024 CASP15 natural RNA comparison R1149-rmsd: R1149: RMSD
Dataset subset: CASP15 natural RNA target R1149 (RhoFold+ 2024 CASP15 natural RNA comparison split)
7.66 Å RMSD
angstrom · lower

Uncertainty: Not yet extracted

Coverage: 1/1 CASP15 natural RNA target scored (Per-cell target count; source retains up to five candidate structures, not five independent evaluation targets.)

Author-reported evaluation · Source checked
Methods, coverage and source

RhoFold+ on RhoFold+ 2024 CASP15 natural RNA comparison R1149-rmsd: R1149: RMSD

Retrospective evaluation of six CASP15 natural single-chain RNA targets; each task retains one named target. Artificially designed targets and RNA-protein complexes excluded. RNA sequence inputs without ions or other molecules. Paper Par13 selects the highest-performing of five candidates; Par51 describes five sampled-MSA runs for RhoFold+, one AlphaFold3 server run yielding five models, and five CASP15 models for the other methods. Figure 2h reports RMSD and sum of Z-scores for GDT-TS and TM-score. Metric-specific winning candidate identity, exact scoring-tool revision and RMSD atom selection remain unextracted. No target pooling or score recomputation.

Aggregation: Source-reported selected candidate for one target; no cross-target aggregation.

RhoFold+ Nature Methods 2024: Source Data Figure 2h · fig2hi!F11; method fig2hi!A11; target fig2hi!F1; Fig. 2h caption; XML Par13–15 and Par51
Configuration: RhoFold+Protocol: RhoFold+ 2024 CASP15 natural RNA comparison R1149-sum_z_gdt_ts_tm_score: R1149: Sum of Z-scores for GDT-TS and TM-score
Dataset subset: CASP15 natural RNA target R1149 (RhoFold+ 2024 CASP15 natural RNA comparison split)
3.75 Casp sum z score
unitless · higher

Uncertainty: Not yet extracted

Coverage: 1/1 CASP15 natural RNA target scored (Per-cell target count; source retains up to five candidate structures, not five independent evaluation targets.)

Author-reported evaluation · Source checked
Methods, coverage and source

RhoFold+ on RhoFold+ 2024 CASP15 natural RNA comparison R1149-sum_z_gdt_ts_tm_score: R1149: Sum of Z-scores for GDT-TS and TM-score

Retrospective evaluation of six CASP15 natural single-chain RNA targets; each task retains one named target. Artificially designed targets and RNA-protein complexes excluded. RNA sequence inputs without ions or other molecules. Paper Par13 selects the highest-performing of five candidates; Par51 describes five sampled-MSA runs for RhoFold+, one AlphaFold3 server run yielding five models, and five CASP15 models for the other methods. Figure 2h reports RMSD and sum of Z-scores for GDT-TS and TM-score. Metric-specific winning candidate identity, exact scoring-tool revision and RMSD atom selection remain unextracted. No target pooling or score recomputation.

Aggregation: Source-reported selected candidate for one target; no cross-target aggregation.

RhoFold+ Nature Methods 2024: Source Data Figure 2h · fig2hi!L11; method fig2hi!A11; target fig2hi!L1; Fig. 2h caption; XML Par13–15 and Par51
Configuration: RhoFold+Protocol: RhoFold+ 2024 CASP15 natural RNA comparison R1156-rmsd: R1156: RMSD
Dataset subset: CASP15 natural RNA target R1156 (RhoFold+ 2024 CASP15 natural RNA comparison split)
17.43 Å RMSD
angstrom · lower

Uncertainty: Not yet extracted

Coverage: 1/1 CASP15 natural RNA target scored (Per-cell target count; source retains up to five candidate structures, not five independent evaluation targets.)

Author-reported evaluation · Source checked
Methods, coverage and source

RhoFold+ on RhoFold+ 2024 CASP15 natural RNA comparison R1156-rmsd: R1156: RMSD

Retrospective evaluation of six CASP15 natural single-chain RNA targets; each task retains one named target. Artificially designed targets and RNA-protein complexes excluded. RNA sequence inputs without ions or other molecules. Paper Par13 selects the highest-performing of five candidates; Par51 describes five sampled-MSA runs for RhoFold+, one AlphaFold3 server run yielding five models, and five CASP15 models for the other methods. Figure 2h reports RMSD and sum of Z-scores for GDT-TS and TM-score. Metric-specific winning candidate identity, exact scoring-tool revision and RMSD atom selection remain unextracted. No target pooling or score recomputation.

Aggregation: Source-reported selected candidate for one target; no cross-target aggregation.

RhoFold+ Nature Methods 2024: Source Data Figure 2h · fig2hi!G11; method fig2hi!A11; target fig2hi!G1; Fig. 2h caption; XML Par13–15 and Par51
Configuration: RhoFold+Protocol: RhoFold+ 2024 CASP15 natural RNA comparison R1156-sum_z_gdt_ts_tm_score: R1156: Sum of Z-scores for GDT-TS and TM-score
Dataset subset: CASP15 natural RNA target R1156 (RhoFold+ 2024 CASP15 natural RNA comparison split)
1.32 Casp sum z score
unitless · higher

Uncertainty: Not yet extracted

Coverage: 1/1 CASP15 natural RNA target scored (Per-cell target count; source retains up to five candidate structures, not five independent evaluation targets.)

Author-reported evaluation · Source checked
Methods, coverage and source

RhoFold+ on RhoFold+ 2024 CASP15 natural RNA comparison R1156-sum_z_gdt_ts_tm_score: R1156: Sum of Z-scores for GDT-TS and TM-score

Retrospective evaluation of six CASP15 natural single-chain RNA targets; each task retains one named target. Artificially designed targets and RNA-protein complexes excluded. RNA sequence inputs without ions or other molecules. Paper Par13 selects the highest-performing of five candidates; Par51 describes five sampled-MSA runs for RhoFold+, one AlphaFold3 server run yielding five models, and five CASP15 models for the other methods. Figure 2h reports RMSD and sum of Z-scores for GDT-TS and TM-score. Metric-specific winning candidate identity, exact scoring-tool revision and RMSD atom selection remain unextracted. No target pooling or score recomputation.

Aggregation: Source-reported selected candidate for one target; no cross-target aggregation.

RhoFold+ Nature Methods 2024: Source Data Figure 2h · fig2hi!M11; method fig2hi!A11; target fig2hi!M1; Fig. 2h caption; XML Par13–15 and Par51

Source checking is not independent reproduction. Release 2026-11-21-c8bea25e5abb.

Use this configuration

How it works, versions and access

Related profile: RhoFold+. This page retains the exact record and its evaluation context.

This configuration

RhoFold+ automated five-candidate sampled-MSA configuration evaluated by its authors.

record
RhoFold+
configuration
Not reported
entity type
Configuration

How it works

How it works

RNA-FM embeddings and MSA representations enter the Rhoformer transformer stack; a geometry-aware invariant-point-attention structure module predicts frames and torsion angles with recycling. The documented inputs are RNA FASTA (A, U, G, C), optionally a supplied MSA; an MSA is otherwise generated by the full workflow. The output consists of three-dimensional PDB models, predicted distograms, secondary structure and per-residue confidence.

Sources (3)ml4bio/RhoFold: README.md; cuhkaih/rhofold: README.md; rhofold: Journal full-text XML · Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata
Versions and reproducibility

RhoFold+; the README links a pretrained checkpoint and the 2024 Nature Methods paper. The paper limits MSA depth to256 sequences during training and default inference. This is alignment depth, not an RNA-length maximum; the latter remains unextracted.

Sources (3)ml4bio/RhoFold: README.md; cuhkaih/rhofold: README.md; rhofold: Journal full-text XML · Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata
Strengths, limitations and unresolved questions

Strengths and limitations

Strengths and considerations

  • Produces secondary and tertiary outputs and supports provided or automatically generated alignments.
    Sources (3)ml4bio/RhoFold: README.md; cuhkaih/rhofold: README.md; rhofold: Journal full-text XML · Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata

Limitations and conditions

  • The README labels sequence-only inference as a lower-accuracy testing mode. Constructing the full MSA databases needs substantial local storage; macOS is not supported by its documented setup.
    Sources (3)ml4bio/RhoFold: README.md; cuhkaih/rhofold: README.md; rhofold: Journal full-text XML · Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata
Profile review details

Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied.

Stable record: catalog-model-rhofold

Specifications

Inputs, training, access and other details

Explanatory profile: limited source coverage · Automated source review, 2026-09-16. Review applies to the cited claims; unresolved fields are listed below. Numerical results retain their own review status.

Inputs, outputs and configuration
PropertyDescription and evidence
Model typeRNA structure predictor with language-model and MSA inputs
Sources (3)ml4bio/RhoFold: README.md; cuhkaih/rhofold: README.md; rhofold: Journal full-text XML · Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata
ArchitectureRNA-FM embeddings and MSA representations enter the Rhoformer transformer stack; a geometry-aware invariant-point-attention structure module predicts frames and torsion angles with recycling.
Sources (3)ml4bio/RhoFold: README.md; cuhkaih/rhofold: README.md; rhofold: Journal full-text XML · Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata
InputsRNA FASTA (A, U, G, C), optionally a supplied MSA; an MSA is otherwise generated by the full workflow.
Sources (3)ml4bio/RhoFold: README.md; cuhkaih/rhofold: README.md; rhofold: Journal full-text XML · Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata
OutputsThree-dimensional PDB models, predicted distograms, secondary structure and per-residue confidence.
Sources (3)ml4bio/RhoFold: README.md; cuhkaih/rhofold: README.md; rhofold: Journal full-text XML · Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata
ParametersThe inspected paper describes RNA-FM, Rhoformer and structure modules without stating a total for the complete selected predictor in those architecture sections. · Not reported in inspected sources
Sources (3)ml4bio/RhoFold: README.md; cuhkaih/rhofold: README.md; rhofold: Journal full-text XML · Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata
Known versionsRhoFold+; the README links a pretrained checkpoint and the 2024 Nature Methods paper.
Sources (3)ml4bio/RhoFold: README.md; cuhkaih/rhofold: README.md; rhofold: Journal full-text XML · Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata
Training dataRNA-FM pretraining uses RNAcentral100. Structure training uses PDB RNA chains selected through BGSU representative sets, with additional self-distillation from RNAStralign/bpRNA-derived sequences.
Sources (3)ml4bio/RhoFold: README.md; cuhkaih/rhofold: README.md; rhofold: Journal full-text XML · Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata
Training cutoffStructural-data selection uses the BGSU representative set dated 2022-04-13. This does not establish a single cutoff for every RNA-FM or MSA resource.
Sources (3)ml4bio/RhoFold: README.md; cuhkaih/rhofold: README.md; rhofold: Journal full-text XML · Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata
Context limitsMSA depth is capped at 256 during documented training and default inference. The reviewed paper does not establish a single RNA-length maximum; MSA depth is a different quantity. · Not reported in inspected sources
Sources (3)ml4bio/RhoFold: README.md; cuhkaih/rhofold: README.md; rhofold: Journal full-text XML · Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata
Weights licenceApache-2.0 declared in the official cuhkaih/rhofold model-card metadata; this is distinct from access conditions for training data.
Sources (3)ml4bio/RhoFold: README.md; cuhkaih/rhofold: README.md; rhofold: Journal full-text XML · Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata
AccessOfficial project documentation and implementation: https://github.com/ml4bio/RhoFold
Sources (3)ml4bio/RhoFold: README.md; cuhkaih/rhofold: README.md; rhofold: Journal full-text XML · Paper: Automated end-to-end platform, Large-scale pretraining dataset, Efficient development of a self-distillation dataset, Feature processing with Rhoformer and Data availability; README.md: Usage; official cuhkaih/rhofold card licence metadata
Code licenceApache-2.0
Sourcesml4bio/RhoFold: LICENSE · LICENSE: licence text

Evidence

Source checking verifies the cited claim or transcription. It does not establish independent reproduction.

Evidence table

Inspect claims, sources and review details

Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.

One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.

1 evidence row matching the loaded filters

Claims, original sources and review scope · Release 2026-11-21-c8bea25e5abb
Property and statementOriginal source and locationReview and provenance
Relationship: family
catalog-model-rhofold
Individual claims
RhoFold+ Nature Methods 2024: Source Data Figure 2h

Original source ↗

fig2hi!A11; Figure 2h; primary XML Par13–15 and Par51

Version: Publisher Source Data Fig. 2, DOI 10.1038/s41592-024-02487-0; retrieved 2026-09-23
Retrieved: 2026-09-23T11:23:43.985240+00:00

source checked

["automated-source-review"] · 2026-09-23

Audit details

Source-backed evaluated identity only; no independent reproduction.

Field: links:family:catalog-model-rhofold

Claim: rhofold-2024-casp15-natural-method-rhofold-plus-catalog-model-rhofold-identity-claim

Source artifact SHA-256: 9d87837d6e72c8dafd265b9427c1cddd0223b3073eea26bcb4c6a90085a1bb04

Hash scope: Hash scope not separately documented; inspect source record

Inspected artifact

Sources and history

Release 2026-11-21-c8bea25e5abb · Record review: source checked

1 source record and release historyDownload this release (gzip)
Technical metadata and record identifiers

Technical metadata for this record is tracked under stable ID rhofold-2024-casp15-natural-method-rhofold-plus.

Source locator: fig2hi!A11; Figure 2h; primary XML Par13–15 and Par51

Related records

How to cite

RhoFold+. Rewire biological model benchmark database, release 2026-11-21-c8bea25e5abb. https://benchmarks.rewirebio.io/database/configuration/rhofold-2024-casp15-natural-method-rhofold-plus/

Release 2026-11-21-c8bea25e5abb. Rankings and values change between releases, so keep the release id in the citation. Cite the original publication as well for any value you reuse: each result links its source.

The compilation and annotations made by Rewire in this release are offered under CC BY 4.0; values and text taken from cited sources stay under the terms of those sources and are not covered by this licence. Data licence

Suggest a correction