GraphClust

GraphClust performs structural clustering of RNA sequences, particularly non-coding RNAs (ncRNAs), to identify sequence–structure similarity for ncRNA annotation and discovery.

Key Features:

• Sequence-structure clustering: Compares and clusters RNAs using both sequence and secondary structure information to capture sequence–structure similarity.
• Linear-time, alignment-free algorithm: Implements a novel linear-time, alignment-free methodology for efficient sequence-structure comparison and clustering.
• High speed: Achieves speedups of several orders of magnitude relative to conventional sequence-structure techniques.
• Scalability: Scales to complete genomic sequences and whole transcriptomes, handling datasets containing hundreds of thousands of sequences.
• Benchmark performance: Benchmarked against known ncRNA datasets with clustering quality comparable to state-of-the-art sequence-structure methods.
• Novel element prediction: Capable of identifying novel structural ncRNAs and predicting local structural elements, including applications to long intergenic non-coding RNAs (lincRNAs).
• Empirical result: Successfully predicted 349 local structural RNA elements in a reported study.

Scientific Applications:

• ncRNA annotation: Facilitates annotation of non-coding RNAs by grouping sequences with similar sequence–structure features.
• Discovery of novel structural ncRNAs: Enables identification of previously uncharacterized structural ncRNAs and local structural motifs.
• lincRNA structural analysis: Predicts local structural elements specific to long intergenic non-coding RNAs (lincRNAs).
• Biological association studies: Supports linking identified structural RNAs and predicted elements to biological processes, including associations reported with the human nervous system.

Methodology:

Uses a novel linear-time, alignment-free algorithm to compare and cluster RNAs by sequence and secondary structure.