OperonSEQer

OperonSEQer predicts operon structures in prokaryotic genomes from short-read RNA-sequencing (RNA-seq) data by identifying co-transcribed adjacent genes.

Key Features:

• Machine learning algorithms: Employs multiple machine learning methods to analyze RNA-seq expression patterns across contiguous prokaryotic genes.
• Statistical analysis (Kruskal-Wallis): Uses a non-parametric Kruskal-Wallis test to assess whether adjacent genes are likely co-expressed from the same transcript.
• Threshold voting system: Implements a threshold voting mechanism to adjust prediction stringency and balance coverage versus accuracy.
• Retraining and hyperparameter tuning: Supports retraining of models and redefinition of hyperparameters using custom datasets.
• Validation with long-read sequencing: Compares predictions to publicly available long-read sequencing data for validation and detection of operon pairs missed by other methods.

Scientific Applications:

• Gene expression regulation: Enables identification of co-expressed gene clusters to inform studies of prokaryotic transcriptional regulation.
• CRISPR-based gene modulation: Guides targeting of contiguous gene clusters for CRISPR-mediated perturbation or engineering.
• Pathogen research: Supports characterization of operons in pathogenic bacteria to investigate virulence factors and potential therapeutic targets.
• Biotechnological engineering: Aids design and optimization of operon architectures for industrial applications such as biofuel production and pharmaceuticals.

Methodology:

Integrates statistical analysis with machine learning, applies the Kruskal-Wallis non-parametric test to RNA-seq expression data to assess co-expression of adjacent genes, refines calls via a threshold voting system, supports model retraining and hyperparameter adjustment, and validates predictions against long-read sequencing data.