PESM

PESM predicts the essentiality of microRNAs (miRNAs) using gradient boosting machines (GBMs) from sequence and structural features to identify miRNAs implicated in biological processes and disease.

Key Features:

• Gradient Boosting Machines (GBMs): Integrates sequence and structural features using GBMs to predict miRNA essentiality.
• Feature Extraction: Extracts sequence and structural features from miRNAs for use as model inputs.
• Cross-Validation: Employs 5-fold cross-validation to validate predictive performance.
• Performance Metrics: Quantifies performance using area under the receiver operating characteristic curve (AUC), F-measure, and accuracy (ACC).
• Comparative Performance: Demonstrates superior performance relative to miES, Gaussian Naive Bayes, and Support Vector Machine with AUC=0.9117, F-measure=0.8572, and ACC=0.8516.
• Feature Importance Analysis: Assesses feature importance and reports that newly incorporated features significantly enhance predictive accuracy.

Scientific Applications:

• Identification of essential miRNAs: Supports identification of miRNAs involved in cell growth and apoptosis.
• Biomarker discovery: Informs selection of miRNAs as potential biomarkers for disease diagnosis and treatment strategies.
• Mechanistic insight: Provides feature-level insights to aid molecular biology studies of miRNA regulatory roles.

Methodology:

Extraction of sequence and structural features from miRNAs, training and prediction with gradient boosting machines (GBMs), evaluation by 5-fold cross-validation using AUC, F-measure, and ACC, and analysis of feature importance.