MicroRNA prediction using a fixed-order Markov model based on the secondary structure pattern.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Predicting microRNAs (miRNAs) is challenging due to precursor diversity and complex enzymatic processes.
• Existing prediction methods often struggle with full-function recognition of mature miRNAs directly from hairpin precursors across different species.
• There is a continuous need for more powerful models that closely mimic biological recognition of miRNA structure.

Purpose of the Study:

• To introduce FOMmiR, a novel miRNA prediction algorithm.
• To develop a model capable of full-function recognition of mature miRNAs directly from hairpin sequences.
• To enhance understanding of biological recognition mechanisms in miRNA maturation.

Main Methods:

• Developed FOMmiR, a fixed-order Markov model utilizing secondary structural patterns.
• Trained and evaluated the model using a dataset of 809 human pre-miRNAs and 6441 human pseudo-miRNA hairpins.
• Validated performance using 5-fold cross-validation and independent test datasets across various species.

Main Results:

• FOMmiR achieved 91% accuracy on the human dataset via 5-fold cross-validation.
• Demonstrated outstanding prediction performance on independent datasets, including vertebrates, Drosophila, worms, viruses, and plants.
• Successfully distinguished miRNA precursors from hairpins and precisely located mature miRNA position and strand.

Conclusions:

• FOMmiR represents a new generation of miRNA prediction algorithms with full-function recognition capabilities.
• The algorithm provides insights into biological recognition, potentially linked to enzyme cleavage mechanisms in miRNA maturation.
• FOMmiR offers a powerful tool for accurate miRNA prediction and analysis across diverse organisms.