biomarker-discovery

Preventing data leakage in biomarker discovery requires strict separation of feature selection and model training within nested cross-validation. This skill audits your study design to identify temporal leakage, patient-level cross-validation errors, and improper feature selection before analysis begins.

Validating a prognostic gene signature requires progressing through a hierarchy from internal nested cross-validation to external replication and prospective testing. This skill enforces decision curve analysis and pre-specified thresholds to confirm diagnostic or prognostic utility beyond mere statistical correlation.

Biomarker study designs fail when the intent—prognostic, predictive, or diagnostic—is not aligned with the clinical question. This skill provides methodological auditing to evaluate biomarker intent and ensure the study design matches the required clinical utility.

Yes, nested cross-validation is essential for clinical and imaging data analysis to ensure methodological integrity. This skill applies a structured reasoning framework across high-dimensional omics, clinical, and imaging data to enforce rigorous validation standards.

Decision curve analysis is needed during biomarker validation to confirm that a model provides clinical utility over default strategies. This skill requires adherence to decision curve analysis alongside pre-specified thresholds to justify diagnostic or prognostic utility.