LncMachine

LncMachine predicts coding potential of long noncoding RNAs (lncRNAs) in plants using alignment-free machine learning to support lncRNA annotation.

Key Features:

• Alignment-free prediction: Performs coding potential prediction without sequence alignment.
• Feature selection: Applies comprehensive feature selection to remove irrelevant or redundant features.
• Machine learning algorithms: Implements Random Forest and supports evaluation and deployment of other machine learning algorithms, including user-provided algorithms in real time.
• Evaluation framework: Uses 10-fold cross-validation to compare algorithm performance.
• Benchmarking: Reported average accuracy of 92.67% on human and mouse datasets and outperformed CPC2, CPAT, and CNIT in prediction accuracy.
• Input formats: Accepts FASTA files or TAB-separated CSV files containing relevant features.
• High-throughput sequencing data: Leverages features derived from high-throughput sequencing data with examples in crop species such as wheat.

Scientific Applications:

• Plant lncRNA annotation: Annotation of long noncoding RNAs in plant genomes, including crop species such as wheat.
• Coding potential assessment: Identification of coding versus noncoding transcripts for lncRNA discovery.
• Cross-species applicability: Application of models to nonplant datasets, as demonstrated on human and mouse data.
• Algorithm benchmarking and development: Comparative evaluation of machine learning methods and deployment of custom algorithms for method development.

Methodology:

Alignment-free coding potential prediction using features derived from high-throughput sequencing, comprehensive feature selection, machine learning classification with Random Forest (and other algorithms), 10-fold cross-validation for performance comparison, and input via FASTA or TAB-separated CSV with optional deployment of user-provided algorithms in real time.