StackRAM

StackRAM identifies N6-methyladenosine (m6A) sites across multiple species using feature extraction, Elastic Net selection, and a stacked ensemble learning framework for accurate m6A site prediction.

Key Features:

• Cross-Species Application: Identifies m6A sites in Saccharomyces cerevisiae, Homo sapiens, and Arabidopsis thaliana.
• Feature Extraction and Fusion: Extracts RNA sequence features using binary encoding, chemical property analysis, nucleotide frequency, k-mer nucleotide frequency, pseudo dinucleotide composition, and position-specific trinucleotide propensity, and fuses these features into a comprehensive dataset.
• Feature Selection with Elastic Net: Applies the Elastic Net algorithm to filter redundant and noisy features and retain an optimal subset for classification.
• Ensemble Learning (Stacked): Employs a stacked ensemble in which base-classifiers generate output probabilities that are combined with the Elastic Net–selected feature subset and input to a Support Vector Machine (SVM) meta-classifier.
• Evaluation and Performance: Demonstrates superior predictive accuracy on the S. cerevisiae jackknife test and achieves 92.30% accuracy for H. sapiens and 87.06% accuracy for A. thaliana on independent test datasets.

Scientific Applications:

• Comparative epitranscriptomics: Enables cross-species identification of m6A sites to support comparative analyses of RNA methylation patterns.
• Gene expression regulation: Facilitates identification of m6A-modified transcripts relevant to studies of gene expression regulation.
• RNA stability analysis: Supports investigation of m6A effects on RNA stability.
• Cellular process investigation: Assists studies of cellular processes influenced by m6A methylation.

Methodology:

Extracts multiple RNA sequence features (binary encoding, chemical property, nucleotide frequency, k-mer frequency, pseudo dinucleotide composition, position-specific trinucleotide propensity), fuses them, applies Elastic Net for feature selection, uses base-classifiers to produce output probabilities that are combined with the selected features in a stacked ensemble, and employs a Support Vector Machine (SVM) as the meta-classifier with evaluation by jackknife and independent tests.