Knotify: An Efficient Parallel Platform for RNA Pseudoknot Prediction Using Syntactic Pattern Recognition.

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Noncoding RNA subsequences are crucial for biological operations but challenging to identify.
• Accurate RNA secondary structure prediction is vital for understanding RNA function.
• The COVID-19 pandemic highlighted the need for efficient RNA analysis tools, particularly for SARS-CoV-2.

Purpose of the Study:

• To develop a novel system for predicting RNA secondary structure patterns, focusing on pseudoknots.
• To leverage syntactic pattern-recognition strategies for RNA secondary structure prediction.
• To address the challenge of identifying biologically realistic RNA folding patterns.

Main Methods:

• Formalized RNA secondary structure prediction as a parsing and optimization problem.
• Introduced a context-free grammar (CFG) for recognizing potential pseudoknot patterns.
• Implemented a novel heuristic for pseudoknot prediction, outperforming free-energy minimization in ambiguity resolution.
• Utilized a brute-force algorithm as an alternative method for pseudoknot detection.

Main Results:

• The methodology successfully predicted core stems of RNA pseudoknots with a 76.4% recall ratio.
• Achieved a F1-score of 0.774 and MCC of 0.543 in discovering all stems of RNA sequences.
• Demonstrated performance speeds 1.31x, 3.45x, and 7.76x faster than three established platforms on a dataset of 262 RNA sequences.
• The system exhibits polynomial-time complexity, with performance gains from parallel implementation.

Conclusions:

• The novel system effectively predicts RNA pseudoknots using syntactic pattern recognition and a heuristic approach.
• The method offers an efficient and performant solution for RNA secondary structure analysis, crucial for understanding viral RNA like SARS-CoV-2.
• Publicly available source code (knotify github repo) facilitates further research and application.