Glossary

CASP-RNA

The RNA-specific track of the Critical Assessment of Structure Prediction experiment, providing a standardized, biennial blind benchmark for comparing the performance of computational RNA structure prediction methods.

Get in touch Learn more

Research scientist tracking AI experiments on laptop, experiment results visible, casual lab environment.

BENCHMARKING RNA STRUCTURE PREDICTION

What is CASP-RNA?

CASP-RNA is the RNA-specific track of the biennial Critical Assessment of Structure Prediction experiment, providing a standardized blind benchmark for evaluating computational methods that predict RNA secondary and tertiary structures.

CASP-RNA is the RNA-focused extension of the global Critical Assessment of Structure Prediction (CASP) experiment. It operates on a biennial cycle where experimentalists provide unpublished, soon-to-be-solved RNA structures as blind prediction targets. Computational groups worldwide then submit predicted models, which are subsequently compared against the experimentally determined structures using standardized metrics like Root Mean Square Deviation (RMSD) and Template Modeling Score (TM-score) to objectively rank method performance.

The experiment evaluates prediction across multiple categories, including de novo modeling, homology modeling, and structure refinement. Unlike protein-centric CASP, CASP-RNA specifically assesses the ability to handle non-canonical base pairing, pseudoknots, and complex tertiary motifs like the A-minor motif. Results are published in a special issue of Proteins: Structure, Function, and Bioinformatics, providing a definitive, community-validated snapshot of the state-of-the-art in RNA computational biology.

BENCHMARK ARCHITECTURE

Key Features of CASP-RNA

The Critical Assessment of Structure Prediction for RNA provides the community with a rigorous, blinded evaluation framework. It defines the gold standard for comparing computational methods against unpublished experimental structures.

Blinded, Prospective Prediction

The core mechanism of CASP-RNA is true blind prediction. Organizers solicit recently solved but unpublished RNA structures from experimentalists. Participants submit models before the experimental coordinates are released, ensuring no accidental overfitting or data leakage occurs. This contrasts sharply with retrospective benchmarks where test sets may overlap with training data.

Biennial

Assessment Cycle

Multi-Category Assessment

CASP-RNA evaluates predictions across distinct structural categories to isolate specific modeling challenges:

RNA-only targets: Isolated riboswitches, ribozymes, and aptamers.
Protein-RNA complexes: Targets requiring accurate modeling of intermolecular interfaces.
RNA-ligand complexes: Structures with small molecule ligands, testing docking accuracy. This stratification prevents methods specialized for one category from dominating the overall ranking.

Quantitative Evaluation Metrics

Predictions are ranked using a suite of complementary metrics to avoid single-score bias:

Root Mean Square Deviation (RMSD): Measures global atomic distance error after optimal superposition.
Template Modeling Score (TM-score): A length-independent metric sensitive to overall topology.
Local Distance Difference Test (lDDT): Evaluates local backbone accuracy, ignoring flexible regions.
Interaction Network Fidelity (INF): Quantifies the correctness of predicted base-pairing and stacking interactions.

Evaluation Metrics

Independent Expert Curation

A panel of structural biologists and computational experts independently analyzes the results. They identify systematic failures, such as the inability to predict pseudoknots or non-canonical base pairs, and publish detailed assessment papers. This human expert layer contextualizes the quantitative scores, distinguishing between a near-native prediction and a model that scores well by chance.

Community-Wide Method Advancement

CASP-RNA functions as a forcing function for innovation. The release of each assessment's results directly catalyzes the next generation of algorithms. For example, the observed failure of pure physics-based methods in early CASP-RNA rounds directly motivated the development of geometric deep learning approaches like RoseTTAFoldNA and AlphaFold 3, which now dominate the leaderboard.

Prediction Confidence Estimation

Modern CASP-RNA assessments evaluate not just structural accuracy but also a model's ability to estimate its own uncertainty. Methods are scored on self-assessment metrics that compare per-residue confidence scores (like pLDDT) against actual local errors. A perfect prediction with overconfident error estimates is penalized, encouraging honest reporting of model reliability for downstream experimental design.

CASP-RNA BENCHMARK

Frequently Asked Questions

Essential questions about the Critical Assessment of Structure Prediction for RNA, the definitive community benchmark for evaluating computational methods in RNA tertiary structure prediction.

CASP-RNA is the RNA-specific track of the Critical Assessment of Structure Prediction (CASP) experiment, a biennial community-wide blind benchmark that evaluates the state-of-the-art in computational RNA structure prediction. The experiment operates on a strict double-blind protocol: experimental structural biologists provide unpublished, soon-to-be-released RNA structures as prediction targets, while participating research groups submit predicted 3D atomic coordinates without any knowledge of the experimental results. An independent assessment panel then evaluates all submissions using standardized metrics including Root Mean Square Deviation (RMSD), Template Modeling Score (TM-score) , and Predicted Local Distance Difference Test (pLDDT). The results are published in a special issue of Proteins: Structure, Function, and Bioinformatics, providing an objective, comparative analysis of method performance across diverse RNA architectures including riboswitches, ribozymes, and viral RNA elements.

Enabling Efficiency, Speed & Accuracy

Intelligent Analysis, Decision & Execution

We build AI systems for teams that need search across company data, workflow automation across tools, or AI features inside products and internal software.

Talk to Us

Search across company data

Give teams answers from docs, tickets, runbooks, and product data with sources and permissions.

Useful when people spend too long searching or get different answers from different systems.

Enterprise searchRAGPermissions

Automate internal workflows

Use AI to route work, draft outputs, trigger actions, and keep approvals and logs in place.

Useful when repetitive work moves across multiple tools and teams.

AI agentsWorkflow automationGovernance

Add AI to products and internal tools

Build assistants, guided actions, or decision support into the software your team or customers already use.

Useful when AI needs to be part of the product, not a separate tool.

AI integrationDecision supportModel routing

CASP-RNA ECOSYSTEM

Related Terms

The CASP-RNA benchmark is defined by its evaluation metrics, historical context, and relationship to other community challenges. These concepts are essential for interpreting competition results.

Root Mean Square Deviation (RMSD)

The standard metric for quantifying the global similarity between a predicted 3D RNA structure and an experimentally determined reference. RMSD is calculated by finding the optimal rigid-body superposition of two structures and measuring the average atomic distance between equivalent atoms.

Units: Ångströms (Å); lower values indicate better predictions
Limitation: Highly sensitive to local outliers; a single misplaced helix can inflate the score
Usage in CASP-RNA: Reported alongside TM-score for a balanced assessment

< 5 Å

Good prediction threshold

Template Modeling Score (TM-score)

A length-independent metric for assessing global structural similarity that addresses RMSD's sensitivity to domain size. TM-score ranges from 0 to 1, where 1.0 indicates a perfect match and values above 0.5 suggest the same overall fold.

Weighting: Uses a distance-dependent weighting scheme that emphasizes global topology over local fluctuations
Formula: Normalized by the square root of sequence length, making scores comparable across RNAs of different sizes
CASP-RNA Role: A primary ranking metric because it is less biased by target length

> 0.5

Same fold threshold

Predicted Local Distance Difference Test (pLDDT)

A per-residue confidence metric output by AlphaFold and related models that estimates the local accuracy of the predicted structure. pLDDT is computed by comparing predicted inter-residue distances to the actual distances in the model.

Scale: 0–100; residues above 90 are modeled with high accuracy, below 50 indicate disorder
Interpretation: Serves as a critical filter for deciding which regions of a prediction to trust for downstream analysis
CASP-RNA Relevance: Evaluators assess whether confidence estimates correlate with actual prediction quality

> 90

High-confidence residue

RNA-Puzzles

A community-wide blind assessment experiment that evaluates the state-of-the-art in RNA tertiary structure prediction. Unlike CASP-RNA's biennial schedule, RNA-Puzzles operates continuously, challenging participants to predict unpublished crystallographic or cryo-EM structures.

Complementarity: RNA-Puzzles focuses exclusively on RNA, while CASP-RNA is a track within the broader protein-centric CASP
Historical Role: Established the foundational benchmarks and community standards that informed CASP-RNA's design
Data Sharing: Solved structures are deposited as public benchmarks for method development

EXPLORE

Leontis-Westhof Classification

A geometric ontology that systematically categorizes RNA base pairs by the interacting edges and glycosidic bond orientation. This classification is essential for annotating the non-Watson-Crick interactions that define RNA tertiary architecture.

Edges: Watson-Crick, Hoogsteen, Sugar
Orientation: Cis or trans relative to the glycosidic bond
CASP-RNA Application: Used to evaluate whether predictions correctly capture the full interaction network, not just the backbone trace

CASP (Historical Context)

The Critical Assessment of Structure Prediction experiment, established in 1994, is the biennial community-wide benchmark that spawned the CASP-RNA track. CASP operates on a blind prediction model: experimentalists provide unpublished structures, and predictors submit models before the structures are released.

Scope: Primarily protein structure prediction; the RNA track was added to address the growing importance of RNA biology
Impact: The dramatic improvements in CASP, notably by AlphaFold2 in CASP14, catalyzed the field of AI-driven structural biology
Format: Predictors receive amino acid or nucleotide sequences and must return atomic coordinates by a strict deadline

EXPLORE

About the author

Prasad Kumkar

CEO & MD, Inference Systems

Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.

His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.

Limited slotsGet a Free AI Consultation

How We Work

Custom AI workflows for your Business

One-fit-all AI don't work for modern businesses. At Inferensys, we aim to understand your business & custom requirements; which we use to define most efficient agentic workflows, the data, and the tools for your business.