MicroRNA identification using linear dimensionality reduction with explicit feature mapping.

Area of Science:

• Bioinformatics
• Molecular Biology
• Genomics

Background:

• microRNAs (miRNAs) are small regulatory RNAs crucial for gene expression in animals and plants.
• Accurate identification of miRNAs is vital for understanding gene regulation.
• Precursor microRNAs (pre-miRNAs) possess characteristic hairpin secondary structures used for identification.

Purpose of the Study:

• To develop and evaluate a novel classifier for identifying precursor microRNAs (pre-miRNAs).
• To assess the performance of the classifier in distinguishing pre-miRNAs from pseudo hairpins and other non-coding RNAs.
• To explore the effectiveness of dimensionality reduction and feature mapping in miRNA classification.

Main Methods:

• Employed a classifier utilizing linear dimensionality reduction combined with explicit feature mapping (miLDR-EM).
• Evaluated classifier performance using a geometric mean (Gm) metric.
• Investigated the impact of using a small number of features, including minimum free energy properties.

Main Results:

• The miLDR-EM classifier achieved a geometric mean of 92.20% with three features and 92.91% with seven features.
• Demonstrated accurate identification of microRNAs using only three minimum free energy properties.
• Showcased the efficacy of explicit feature mapping as an alternative to kernel methods for large datasets.

Conclusions:

• Linear dimensionality reduction with explicit feature mapping (miLDR-EM) provides high performance for miRNA classification.
• Explicit feature mapping is a viable alternative to kernel-based methods for large-scale datasets.
• Accurate microRNA identification is achievable using a minimal set of features, particularly those related to minimum free energy.