ASmiR

ASmiR predicts abiotic stress-specific microRNAs (miRNAs) in plants using machine learning to identify miRNAs responsive to cold, drought, heat, and salt stresses.

Key Features:

• Machine Learning-Based Prediction Model: Implements Support Vector Machines (SVM) for miRNA prediction and reports performance superior to various deep learning models.
• Stress-Specific Classification: Performs stress-specific predictions for four abiotic stresses: cold, drought, heat, and salt.
• Feature Representation: Represents miRNA sequences using pseudo K-tuple nucleotide compositional features with k-mer sizes from 1 to 5.
• Feature Selection: Applies a feature selection strategy to identify and use the most predictive features for model efficiency and accuracy.
• Performance Metrics: Reports cross-validation areas under precision-recall curves (AUPRC) of 90.15%, 90.09%, 87.71%, and 89.25% for cold, drought, heat, and salt respectively, and independent-dataset accuracies of 84.57%, 80.62%, 80.38%, and 82.78% for the same stresses.

Scientific Applications:

• Crop Breeding Programs: Identification of abiotic stress-responsive miRNAs to inform breeding of cultivars with enhanced resistance to environmental stresses.
• Plant Stress Physiology Research: Provides insights into molecular mechanisms of plant responses to abiotic stresses through stress-specific miRNA prediction.

Methodology:

Transforms miRNA sequences into pseudo K-tuple nucleotide compositional features (k = 1–5), applies feature selection, and trains and validates a Support Vector Machine (SVM) model on datasets representing cold, drought, heat, and salt stresses.