ASLncR

ASLncR predicts abiotic stress-responsive long non-coding RNAs (lncRNAs) in plants to identify lncRNAs involved in responses to drought, salinity, and extreme temperatures.

Key Features:

• Target: Predicts plant lncRNAs responsive to abiotic stresses including drought, salinity, and extreme temperatures.
• Task: Implements a binary classification framework to distinguish stress-responsive versus non-responsive lncRNA sequences.
• Data: Training dataset comprises 263 sequences per class and an independent test set comprises 101 sequences per class.
• Feature representation: Uses Kmer features of sizes 1 to 6 to numerically represent lncRNA sequences.
• Feature selection: Applies four different feature selection strategies to identify significant features for model training.
• Algorithm evaluation: Evaluates seven learning algorithms and selects the best-performing model.
• Selected model and cross-validation performance: Support Vector Machine (SVM) achieved 5-fold cross-validation accuracy of 68.84% with AU-ROC 72.78% and AU-PRC 75.86%.
• Independent test performance: SVM achieved 76.23% accuracy with AU-ROC 87.71% and AU-PRC 88.49% on the independent test set.

Scientific Applications:

• Identification of stress-responsive lncRNAs: Prioritizes lncRNA candidates implicated in plant abiotic stress responses for downstream study.
• Crop improvement research: Provides candidate lncRNAs to inform breeding and resilience-focused research in crops.
• Functional validation prioritization: Ranks lncRNA candidates for experimental validation of roles in drought, salinity, and temperature responses.

Methodology:

Machine learning–based binary classification using Kmer features (k=1–6); training on 263 sequences per class and testing on 101 sequences per class; four feature selection strategies; evaluation of seven learning algorithms; model selection based on 5-fold cross-validation and independent test metrics, with SVM selected as best-performing model.