MSclassifier

MSclassifier performs binary classification of biomedical datasets by augmenting training data with median-supplement-derived samples to improve model balance and classification performance for biomedical decision-making.

Key Features:

• Median-Supplement Approach: Estimates medians of dataset features to generate supplementary data that augment the training set and enhance balance and robustness.
• Binary Classification Capability: Focuses on binary decision-making tasks to distinguish between two outcome classes in biomedical applications.
• Evaluation Methods: Uses independent sample tests and cross-validation to assess classification performance.

Scientific Applications:

• HER2 Receptor Expression Status Phenotyping: Applied to breast cancer data to determine HER2 receptor expression status, achieving 90.30% classification rate versus the best performing tool at 89.83% (p=8.62e-3).
• Protein Subcellular Localization Prediction: Applied to predict plasma membrane and nuclear localization, achieving 93.42% classification performance comparable to another leading tool at 93.19% (p=0.06) and outperforming Naive Bayes-based tools.

Methodology:

Calculating feature medians to generate supplementary data (feature median estimation), integrating supplementary data into the training set (data augmentation), constructing classification models using the augmented dataset optimized for binary tasks (model building), and evaluating performance via independent sample tests and cross-validation.