miPred

miPred classifies genuine microRNA precursors (pre-miRNAs) versus pseudo pre-miRNAs to distinguish functional miRNA precursors for studies of miRNA-mediated gene regulation.

Key Features:

• Hybrid Feature Integration: Combines local contiguous structure-sequence composition, minimum free energy (MFE) of secondary structures, and P-values from randomization tests to distinguish real and pseudo pre-miRNAs.
• Machine Learning Algorithm - Random Forest (RF): Employs a Random Forest (RF) classifier that outperforms the Triplet-SVM-classifier with nearly a 10% increase in total accuracy.
• High Specificity and Sensitivity: Reports a specificity of 98.21% and a sensitivity of 95.09% for pre-miRNA classification.

Scientific Applications:

• miRNA discovery: Identification of genuine pre-miRNAs in genomic and transcriptomic datasets to expand miRNA catalogs.
• Gene regulation studies: Enabling analysis of miRNA-mediated regulation by distinguishing functional miRNA precursors from structural mimics.
• Disease research: Supporting investigation of miRNA roles in diseases such as cancer by providing high-confidence pre-miRNA predictions.
• Therapeutic research: Informing development of targeted therapeutic strategies that involve validated miRNA precursors.
• Large-scale and functional genomics: Facilitating large-scale miRNA studies and functional genomics analyses through high-accuracy precursor classification.

Methodology:

Integrates local contiguous structure-sequence composition with minimum free energy (MFE) calculations and P-values from randomization tests, and classifies sequences using the Random Forest (RF) algorithm.