NmRF

NmRF predicts 2'-O-methylation modification sites across miRNA, tRNA, and mRNA to identify post-transcriptional loci of 2'-O-methyltransferase-catalyzed methylation.

Key Features:

• Predictive scope: Identifies 2'-O-methylation sites across multiple RNA types including miRNA, tRNA, and mRNA.
• Modification chemistry: Targets 2'-O-methylation, the enzymatic replacement of the hydrogen on the 2'-hydroxyl group with a methyl group catalyzed by 2'-O-methyltransferase enzymes.
• Feature engineering: Uses an optimal mixed (hybrid) feature set derived by an incremental feature selection strategy.
• Feature selection algorithm: Employs a light gradient boosting algorithm with incremental feature selection to derive the optimal hybrid feature set.
• Classifier: Constructs the final predictor using a random forest classifier trained on the optimal mixed feature set.
• Validation: Performance assessed by 10-fold cross-validation and independent tests.
• Performance metrics: Reported accuracy of 89.069% (Homo sapiens) and 93.885% (Saccharomyces cerevisiae) with corresponding AUC values of 0.9498 and 0.9832.
• Benchmarking: Demonstrates superior accuracy compared to existing tools in reported evaluations.

Scientific Applications:

• Post-transcriptional modification mapping: Prediction of 2'-O-methylation sites to support mapping of post-transcriptional modifications on RNA molecules.
• Functional mechanism exploration: Facilitates investigation of the functional mechanisms underlying 2'-O-methylation modifications.
• Disease-related studies: Supports analyses linking 2'-O-methylation to disease mechanisms.
• Cross-species analysis: Enables comparative prediction and evaluation across species, exemplified by Homo sapiens and Saccharomyces cerevisiae results.

Methodology:

An incremental feature selection strategy using a light gradient boosting algorithm derives an optimal hybrid feature set, which is then used to train a random forest classifier; performance is evaluated by 10-fold cross-validation and independent tests.