Vernal

Vernal mines recurrent 3D RNA motifs characterized by networks of base-pair interactions by representing RNA structures as graphs and applying graph representation learning and clustering to detect flexible and variable motif classes.

Key Features:

• Graph Representation Learning: Represents RNA structures as continuous graph embeddings to capture flexibility and structural variability.
• Clustering Methods: Applies clustering algorithms to group similar graph-derived regions and identify recurrent motif classes.
• Node Similarity Functions: Implements functions to assess similarity between nodes representing base pairs and interactions.
• Motif Construction Algorithms: Reconstructs motif models from clustered graph regions to define motif instances.
• Relaxed Structural Constraints: Relaxes traditional constraints and narrow search spaces to accommodate variable and fuzzy motif architectures.

Scientific Applications:

• RNA structural biology: Identification and characterization of recurrent 3D motifs formed by networks of base-pair interactions to inform structure–function relationships.
• Motif discovery and classification: Detection and classification of known and novel motif classes via graph-based representations and clustering.
• RNA network analysis: Analysis of base-pair interaction networks to explore motif diversity and structural variability.

Methodology:

Frames motif mining as a graph representation learning and clustering task using continuous graph embeddings, node similarity functions, clustering algorithms, and motif-construction algorithms, with configurable motif flexibility, abundance, and size and relaxed structural constraints to capture variability.