Leontis-Westhof Classification

The Leontis-Westhof system defines 12 distinct base-pairing families by combining three interacting edges with two relative strand orientations.

• Edges: Watson-Crick, Hoogsteen, Sugar
• Orientations: cis or trans glycosidic bonds
• Notation: cWW (cis Watson-Crick/Watson-Crick), tWH (trans Watson-Crick/Hoogsteen), etc.
• Scope: Covers both canonical and non-canonical pairs, including base-backbone interactions

Each nucleotide presents three distinct edges for hydrogen bonding:

• Watson-Crick Edge: The standard base-pairing face used in canonical A-U and G-C pairs; involves atoms O2, N3, N4 for cytosine and N1, N6 for adenine
• Hoogsteen Edge: The face involving atoms N7 and O6/N6; accessible in the major groove; enables triplex formation and G-quadruplex tetrads
• Sugar Edge: The 2'-hydroxyl and sugar ring atoms; critical for A-minor motifs and ribose zipper interactions

The relative orientation of the glycosidic bonds determines the cis/trans designation:

• Cis: Both glycosidic bonds point toward the same side of the base-pair plane; typical of canonical Watson-Crick pairs (cWW)
• Trans: Glycosidic bonds point to opposite sides; common in Hoogsteen pairings and parallel-stranded interactions
• Significance: Orientation dictates backbone geometry and helix handedness, directly impacting tertiary structure prediction

Automated tools classify base pairs in RNA 3D structures using geometric criteria:

• MC-Annotate: Pioneering tool that parses PDB files and assigns Leontis-Westhof classifications based on hydrogen bond patterns and planar geometry
• FR3D (Find RNA 3D): Web server and MATLAB toolkit for exhaustive annotation, including base-stacking and base-phosphate interactions
• DSSR: Integrated component of the 3DNA suite; provides standardized Leontis-Westhof annotations alongside block representations
• Usage: Essential for building non-redundant motif libraries like the RNA 3D Motif Atlas

Leontis-Westhof classification is the foundation for cataloging recurrent 3D motifs:

• A-Minor Motif: Classified by the insertion of an adenine's sugar edge into the minor groove of a cWW pair; types I, II, and III defined by interaction geometry
• Kink-Turn (K-Turn): A sharp bend stabilized by tandem sheared G-A pairs (tSH) and an A-minor interaction
• Ribose Zipper: Two consecutive sugar-edge/sugar-edge interactions (cSS or tSS) that tightly pack two RNA strands
• Loop E Motif: A symmetric internal loop featuring cross-strand purine stacking and specific tWH pairs

Modern RNA structure prediction models implicitly or explicitly leverage Leontis-Westhof geometry:

• AlphaFold 3: Predicts all-atom coordinates that must satisfy the geometric constraints of non-canonical pairs, validated against Leontis-Westhof annotations
• RoseTTAFoldNA: Its three-track architecture must capture the distinct spatial signatures of cWW, tSH, and sugar-edge interactions
• Training Data: Curated datasets like RNA3DB use Leontis-Westhof classifications to ensure balanced representation of non-canonical interactions
• Evaluation: Predicted structures are assessed for correct base-pair geometry, not just RMSD

The Leontis-Westhof system classifies base pairs by three interacting edges:

• Watson-Crick edge: The canonical hydrogen-bonding face
• Hoogsteen edge: The purine C6/N7 face or pyrimidine C4/N3 face
• Sugar edge: The 2'-OH and sugar ring face

Each pair is further defined by glycosidic bond orientation (cis or trans), yielding 12 distinct geometric families. This ontology captures non-canonical pairs like the G·U wobble and A-minor motifs that standard dot-bracket notation cannot represent.

Beyond Watson-Crick A-U and G-C pairs, RNA structures contain diverse non-canonical interactions critical for tertiary folding:

• G·U wobble: A cis Watson-Crick/Watson-Crick pair with shifted hydrogen bonding
• A·G sheared: A trans Hoogsteen/Sugar edge pair common in GNRA tetraloops
• U·U cis Watson-Crick/Watson-Crick: Stabilizes internal loops and bulges
• A-minor (Type I/II): Adenine sugar edge inserting into the minor groove of a receptor helix

These interactions are systematically cataloged in the NCIR database and are essential for riboswitch and ribozyme function.

The Leontis-Westhof classification enables precise annotation of recurrent 3D motifs—conserved structural building blocks:

• Kink-turn (K-turn): An asymmetric internal loop inducing a sharp ~120° bend, stabilized by tandem sheared G·A pairs
• C-loop: A motif with a characteristic non-canonical pairing pattern that widens the major groove
• Sarcin-ricin loop: A bulged-G motif with a distinct geometry recognized by ribotoxins
• Tetraloop-receptor: Docking interactions where GNRA tetraloops bind specific minor groove receptors

Tools like FR3D and RNAMotif use this ontology to search and compare these motifs across structures.

Two primary software tools implement the Leontis-Westhof classification for automated annotation:

• FR3D (Find RNA 3D): Developed by the Leontis and Zirbel labs, FR3D identifies all base pairs and stacking interactions in a PDB file, outputting geometric isostericity and interaction classifications. It powers the RNA 3D Motif Atlas.
• RNAView: Generates 2D topology diagrams from 3D structures using the Leontis-Westhof nomenclature, producing publication-quality secondary structure representations that include non-canonical pairs.

Both tools are integrated into the PDB and NDB annotation pipelines.

An isostericity matrix quantifies how geometrically interchangeable two base pairs are within a given structural context:

• Base pairs with low isostericity can substitute for each other without disrupting backbone geometry
• High isostericity substitutions require backbone adjustments and are rarely observed in homologous structures
• Matrices are organized by Leontis-Westhof geometric family and are used to evaluate sequence-structure covariation in RNA alignments

This concept is fundamental for RNA 3D homology modeling and understanding evolutionary constraints on non-coding RNA sequences.

A comprehensive, hierarchically organized database of recurrent RNA 3D internal and hairpin loop motifs extracted from all available experimental structures:

• Each motif instance is annotated with full Leontis-Westhof base pair classifications
• Motifs are clustered by geometric similarity into families and subfamilies
• Provides loop isostericity data for evaluating sequence variants
• Used as a fragment library for RNA 3D structure prediction and design

The Atlas is maintained by the Bowling Green State University RNA group and is a core resource for the RNA structural biology community.