CLCL

CLCL identifies non-coding RNA (ncRNA) elements within the 3′ untranslated regions (3′-UTRs) of Drosophila melanogaster to characterize RNA structural elements involved in post-transcriptional gene regulation.

Key Features:

• RNA structural clustering: Employs an RNA structural clustering pipeline that accounts for length-dependent distributions of structural similarity measures.
• Structural-similarity-based clustering: Clusters ncRNA elements based on structural similarities to group candidate structural RNAs.
• Benchmarking against Rfam: Demonstrates over a 10% performance gain relative to traditional hierarchical clustering when benchmarked against Rfam.
• RNAz validation: Uses RNAz to predict which clusters are likely true RNA structural elements, with 91.3% of identified clusters predicted as true in the applied dataset.
• Functional inference: Enables inference of co-expression patterns among genes from clustered ncRNA elements.
• Recovery of known families: Recovers known ncRNA families such as the histone ncRNA family.

Scientific Applications:

• ncRNA discovery in 3′-UTRs: Identification of 184 ncRNA clusters within Drosophila melanogaster 3′-UTRs.
• Structural ncRNA candidate prioritization: Prioritizes candidate structural ncRNAs for experimental follow-up using RNAz predictions.
• Expression and localization studies: Links clustered ncRNAs to tissue- or sex-specific expression and subcellular localization, including a male-specific expression cluster and 'cup' or 'comet' localization patterns in testis observed by in situ hybridization.
• Algorithm evaluation: Provides a framework for comparative assessment of structural clustering methods versus hierarchical clustering using Rfam benchmarks.
• Functional inference of gene interactions: Supports inference of potential functional interactions among genes through co-expression patterns of clustered ncRNAs.

Methodology:

Clusters ncRNA elements based on structural similarities using an RNA structural clustering pipeline that accounts for length-dependent distributions of structural similarity measures, benchmarks performance against Rfam, and assesses clusters with RNAz.