cmfinder

cmfinder predicts RNA motifs in unaligned nucleotide sequences and generates covariance models to characterize conserved RNA secondary structures.

Key Features:

• Expectation Maximization Algorithm: Employs an expectation maximization statistical approach to iteratively refine RNA motif predictions.
• Covariance Models: Uses covariance models to describe RNA motifs and capture sequence-structure covariation.
• Integration of Multiple Techniques: Combines mutual information-based methods with folding energy-based approaches within a Bayesian framework for RNA secondary-structure prediction.
• Robustness and Scalability: Demonstrates performance across datasets with varying sequence similarity, including sequences with long flanking regions and unrelated sequences, and is applicable to large-scale analyses.
• High Accuracy: Reported testing on 19 known non-coding RNA (ncRNA) families achieved an average per-base-pair accuracy of 79%, compared with at most 60% for alternative methods.
• Probabilistic Model for Homology Search: Produces probabilistic models that can be used directly for homology searches and iterative refinement of structural models to identify homologs in deeply diverged species.

Scientific Applications:

• ncRNA discovery and characterization: Predicts conserved RNA motifs to support identification and structural characterization of non-coding RNAs.
• Comparative genomics: Identifies conserved structural elements across diverse species to aid evolutionary analyses.
• Functional annotation: Supports functional annotation of ncRNAs by providing structural models and homology-based evidence.
• Homology detection: Enables detection of homologous RNAs in deeply diverged species via probabilistic covariance models.

Methodology:

Applies expectation maximization using covariance models, integrates mutual information and folding-energy methods within a Bayesian framework, and outputs probabilistic models for iterative homology search and model refinement.