miRFam: an effective automatic miRNA classification method based on n-grams and a multiclass SVM

Area of Science:

• Bioinformatics
• Genomics
• Molecular Biology

Background:

• MicroRNAs (miRNAs) regulate gene expression post-transcriptionally.
• Organizing thousands of identified miRNAs into homologous families is crucial for understanding their functions.
• Existing methods for miRNA classification can be labor-intensive and require manual refinement.

Purpose of the Study:

• To develop an automated and accurate method for classifying newly discovered microRNAs into their corresponding families.
• To leverage supervised learning techniques for miRNA gene classification.
• To utilize primary sequence information for efficient miRNA family assignment.

Main Methods:

• Proposed miRFam, a method employing multiclass Support Vector Machines (SVM).
• Utilized n-grams for feature extraction from precursor miRNA sequences.
• Trained the SVM classifier using known miRNA family data from miRBase.

Main Results:

• Achieved high classification accuracy, reaching approximately 98% for datasets with over 300 families (each with at least 5 members).
• Demonstrated robust performance even with a large, imbalanced dataset (miRBase15), achieving 90% accuracy.
• Showcased machine learning as a more general and effective approach compared to sequence alignment and manual modification.

Conclusions:

• miRFam is a suitable automated tool for microRNA family classification.
• It serves as a valuable supplementary method to existing alignment-based classification techniques for small non-coding RNAs (sncRNAs).
• The method's reliance on primary sequence information makes it broadly applicable.