Glossary

Dot-Bracket Notation

A standard string representation of RNA secondary structure where matching parentheses denote canonical base pairs and dots represent unpaired nucleotides, used as a training target for deep learning models.

Get in touch Learn more

ML engineer managing model training cluster on laptop, GPU utilization visible, technical deep learning setup.

RNA STRUCTURE REPRESENTATION

What is Dot-Bracket Notation?

Dot-bracket notation is the standard string encoding for RNA secondary structure, where matching parentheses denote canonical base pairs and dots represent unpaired nucleotides, serving as a primary training target for deep learning models.

Dot-bracket notation is a one-dimensional string representation that encodes the secondary structure of an RNA molecule. In this format, each character corresponds to a nucleotide in the primary sequence: a ( represents a base paired with a downstream partner, a ) represents a base paired with an upstream partner, and a . denotes an unpaired nucleotide in a loop, bulge, or junction region. The notation captures the nested, non-crossing base pairs that define canonical secondary structure elements such as stems and hairpin loops.

This format serves as the standard output target for secondary structure prediction algorithms, from thermodynamic minimum free energy methods to modern deep learning models like SPOT-RNA and UFold. Because dot-bracket strings are inherently tree-structured and cannot represent pseudoknots (crossing base pairs), extended formats such as WUSS notation or connectivity tables are required for tertiary motifs. The simplicity and direct sequence alignment of dot-bracket notation make it the preferred format for training RNA language models and evaluating prediction accuracy using metrics like F1-score and Matthews correlation coefficient.

RNA Structure Representation

Key Characteristics of Dot-Bracket Notation

Dot-bracket notation is the standard string representation for RNA secondary structure, encoding base-pairing interactions and unpaired regions in a compact, machine-readable format essential for training deep learning models.

Core Encoding Rules

The notation uses three characters to represent nucleotide states:

. (dot): Unpaired nucleotide (loop, bulge, or dangling end)
( (open): Nucleotide paired with a downstream partner
) (close): Nucleotide paired with an upstream partner

Pairs must be balanced and nested—each opening parenthesis has exactly one matching closing parenthesis at the corresponding position.

Training Target for Deep Learning

Dot-bracket strings serve as the ground-truth label for supervised RNA secondary structure prediction models. Neural networks are trained to map raw sequence tokens to per-nucleotide pairing probabilities, with the dot-bracket string as the discrete target.

Models like SPOT-RNA and UFold output a dot-bracket prediction
Loss functions compare predicted vs. true bracket positions
Enables direct benchmarking against thermodynamic methods

Limitations: Pseudoknots

Standard dot-bracket notation cannot represent pseudoknots—non-nested base pairs where nucleotides within a loop pair with nucleotides outside that loop.

Pseudoknots violate the context-free grammar assumption
Extended formats like extended dot-bracket use additional bracket types ([], {}, <>) to encode crossing pairs
Models targeting pseudoknot prediction require these richer representations

Vienna Format Convention

The widely adopted Vienna format pairs a dot-bracket string with the corresponding RNA sequence and a free energy value:

code
>sequence_name
AUCGAUCG
((....))
(-12.40)

This format is the standard output of tools like RNAfold and mfold, and the expected input for many downstream analysis pipelines.

Ensemble Representation

Beyond single structures, dot-bracket notation supports ensemble representations using additional characters:

, (comma): Nucleotide unpaired in all suboptimal structures
| (pipe): Nucleotide paired but with varying partners
{ }: Weakly paired regions

These extended characters capture structural ambiguity and are used in partition function outputs.

Canonical Base Pairs Only

Standard dot-bracket notation implicitly represents only canonical Watson-Crick (A-U, G-C) and wobble (G-U) base pairs. Non-canonical interactions—such as Hoogsteen pairs, A-minor motifs, or base triples—are not captured.

Requires complementary annotation formats like Leontis-Westhof classification for full 3D interaction mapping
Limits the notation's utility for tertiary structure description

RNA STRUCTURE REPRESENTATION

Frequently Asked Questions

Common questions about dot-bracket notation, the standard string representation for RNA secondary structure used as training targets in deep learning models.

Dot-bracket notation is a one-dimensional string representation of RNA secondary structure where matching parentheses ( and ) denote canonical base pairs, and dots . represent unpaired nucleotides. The notation captures the nested topology of base-pairing interactions along the linear RNA sequence. For example, the string (((...))) represents a simple stem-loop: three consecutive base pairs forming a helix, followed by three unpaired loop nucleotides, closed by three complementary base pairs. Each opening parenthesis at position i must have a corresponding closing parenthesis at position j, indicating that nucleotide i pairs with nucleotide j. This format is the standard output of minimum free energy (MFE) prediction algorithms and serves as the primary training target for deep learning models like SPOT-RNA and UFold that predict RNA secondary structure directly from sequence.

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

REPRESENTATION COMPARISON

Dot-Bracket vs. Other Structure Representations

Comparison of RNA secondary structure representation formats used as training targets or outputs for deep learning models.

Feature	Dot-Bracket Notation	Adjacency Matrix	Contact Map
Data Structure	1D string	2D binary matrix	2D distance/heatmap matrix
Human Readability
Encodes Pseudoknots
Memory Complexity	O(N)	O(N²)	O(N²)
Standard for MFE Algorithms
Native Loss Function Compatibility	Cross-entropy (per-position)	Binary cross-entropy	MSE or MAE
Used in AlphaFold 3 Training
Encodes Base Pair Type

ESSENTIAL CONCEPTS

Related Terms

Core computational and experimental concepts that intersect with dot-bracket notation in RNA structure prediction workflows.

RNA Secondary Structure Prediction

The computational task of determining the set of base pairs formed by hydrogen bonding within a single RNA strand. The output is canonically represented in dot-bracket notation, where matching parentheses denote paired bases and dots represent unpaired nucleotides. Modern deep learning models are trained to predict this string directly from sequence, treating it as a sequence-to-sequence translation problem.

>90%

Accuracy on short sequences

Minimum Free Energy (MFE)

The thermodynamic principle that predicts the single most stable RNA secondary structure by minimizing the sum of empirically derived loop and stacking energy parameters. The Turner Energy Model provides the nearest-neighbor parameters for this calculation. The resulting MFE structure is output in dot-bracket notation and often serves as a baseline or training target for deep learning models.

Pseudoknot Prediction

The specific computational challenge of identifying pseudoknots, a tertiary structural motif where bases within a loop pair with bases outside that loop. Standard dot-bracket notation cannot represent pseudoknots because they introduce crossing base pairs that break the nested parenthesis grammar. Extended notations like WUSS or BPSEQ formats are required to encode these non-nested interactions.

SHAPE Reactivity

A chemical probing method that acylates the 2'-hydroxyl of flexible nucleotides, providing per-nucleotide data that correlates with local structural dynamics. SHAPE reactivity values are integrated as pseudo-energy constraints into folding algorithms to improve dot-bracket structure prediction accuracy. Deep learning models increasingly incorporate SHAPE data as an auxiliary input channel alongside the RNA sequence.

Covariance Model

A probabilistic model, typically a stochastic context-free grammar (SCFG), that captures both sequence conservation and correlated base-pair mutations within an RNA family. Covariance models improve homology-based structure prediction by leveraging evolutionary information. The output consensus structure is expressed in dot-bracket notation, and these models underpin databases like Rfam.

End-to-End Learning

A deep learning strategy where a single model directly maps raw RNA sequence to 3D atomic coordinates without relying on separate secondary structure or potential energy subroutines. Models like AlphaFold 3 and RoseTTAFoldNA implicitly learn secondary structure representations, including dot-bracket-like internal states, as an intermediate latent feature rather than an explicit output.

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.