Gene Expression Data Analyzer

Gene Expression Data Analyzer analyzes differential gene expression from cancer microarray studies and identifies predictive biomarkers for class prediction and chemotherapy outcome prediction.

Key Features:

• Maximum Difference Subset (MDSS) Algorithm: The Maximum Difference Subset (MDSS) algorithm integrates classification algorithms, classical statistics, and machine learning and dynamically determines the critical significance threshold (alpha or P-value).
• Reduction of False Positives: A jackknife step refines the predictive gene set by removing genes with low combined predictive utility, reducing false positives and increasing external validity.
• High External Validity: Minimizes dependence on arbitrary study design choices such as sample inclusion or exclusion to improve external validity across studies.
• Predictive Utility in Clinical Applications: Identifies biomarkers meeting both statistical significance and predictive utility criteria, including prediction of response to anthracycline-cytarabine therapy in acute myeloid leukemia.
• Flexibility with Test and Classifier Operators: Compatible with any test and classifier operator pair to support varied analytical configurations.

Scientific Applications:

• Biomarker Discovery in Oncology: Discovers dysregulated genes as biomarkers from cancer microarray datasets.
• Class Prediction and Personalized Treatment: Performs class prediction to inform personalized medicine approaches, including predicting chemotherapy outcomes in acute myeloid leukemia.

Methodology:

Computational methods include the Maximum Difference Subset (MDSS) algorithm with dataset-specific learning of significance level (alpha/P-value), jackknife resampling to remove low-utility genes, hierarchical clustering, t-, F-, and Z-tests, and machine learning classifiers for class discovery and prediction, accepting arbitrary test and classifier operator pairs.