VSOLassoBag

VSOLassoBag applies the Least Absolute Shrinkage and Selection Operator (LASSO) with a bagging ensemble to perform stable variable selection for biomarker discovery in high-dimensional omics datasets.

Key Features:

• Ensemble learning: Incorporates a bagging ensemble of LASSO models to improve variable selection stability in high-dimensional, low-sample-size omics data.
• Variable selection and stability: Aggregates and votes on variables across multiple LASSO models and applies either a parametric method or an inflection point search method to determine candidate biomarkers and mitigate overfitting.
• Performance and validation: Validated on simulation datasets and real-world datasets for marker identification in case-control binary classification and prognosis prediction while selecting fewer features.
• Implementation and computation: Implemented as an R package with multithreading computing configurations to support computational efficiency.

Scientific Applications:

• Biomarker discovery: Selection of reliable biomarkers from high-dimensional omics data for translational research.
• Prognosis prediction: Identification of features predictive of clinical prognosis.
• Case-control classification: Marker selection for binary disease versus control classification tasks.
• Translational biomedical studies: Application to clinical studies requiring interpretable and stable feature selection from omics datasets.

Methodology:

Fits multiple LASSO models within a bagging framework, aggregates and votes on selected variables, and selects final variables using a parametric method or an inflection point search method.