RNA search with decision trees and partial covariance models.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Searching genomic databases for RNA families is computationally intensive.
• Existing methods often require extensive computational resources.

Purpose of the Study:

• To explore the use of partial covariance models for efficient RNA family identification.
• To develop a framework for optimizing the search process and reducing computational time.

Main Methods:

• Utilizing contiguous subranges of RNA family multiple alignment columns to form partial models.
• Implementing a binary decision-tree framework to guide the application of partial models and set score thresholds.
• Employing computational intelligence methods for complex decision tree selection.
• Evaluating sequences using a full covariance model after initial filtering by partial models.

Main Results:

• Achieved execution times of 0.066-0.268 relative to the full covariance model alone across seven RNA families.
• Successfully identified at least 95% of known sequences for two families and 100% for five families.
• Maintained a false alarm rate comparable to or lower than the full covariance model alone.

Conclusions:

• Partial covariance models offer a computationally efficient approach for RNA family searches in genomic databases.
• The decision-tree framework effectively balances speed and accuracy in sequence identification.
• This method provides a significant speedup without compromising the sensitivity or specificity of RNA family detection.