Related Experiment Video
Updated: May 16, 2026
DeepOmicsAE: Representing Signaling Modules in Alzheimer's Disease with Deep Learning Analysis of Proteomics, Metabolomics, and Clinical Data
Published on: December 15, 2023
Exploiting the noise: improving biomarkers with ensembles of data analysis methodologies
Maud Hw Starmans1, Melania Pintilie2, Thomas John3
1Informatics and Biocomputing Platform, Ontario Institute for Cancer Research, Toronto, ON, M5G 0A3, Canada ; Department of Radiation Oncology (Maastro), GROW-School for Oncology and Developmental Biology, Maastricht University Medical Center, Maastricht, The Netherlands.
Technical issues in data analysis hinder biomarker validation for non-small cell lung cancer (NSCLC) treatments. A new method improves biomarker robustness by addressing this sensitivity.
Area of Science:
- Oncology
- Biomarker Discovery
- Personalized Medicine
Background:
- Personalized medicine relies on reproducible biomarkers for effective treatment selection.
- Poor validation and clinical uptake of molecular signatures remain significant challenges.
- This study investigates technical reasons for failed biomarker validation in non-small cell lung cancer (NSCLC).
Purpose of the Study:
- To validate two published prognostic multi-gene biomarkers for NSCLC in an independent cohort.
- To systematically assess the impact of technical factors on biomarker validation success.
- To identify and address data analysis challenges affecting biomarker reliability.
Main Methods:
- Evaluation of two NSCLC prognostic multi-gene biomarkers in a 442-patient dataset.
- Systematic assessment of technical factors influencing validation outcomes.
- Analysis of 24 pre-processing techniques to determine impact on biomarker performance.
Main Results:
- Both biomarkers validated successfully in the independent cohort (P < 0.05).
- Biomarkers stratified patients by stage (II and IB), despite underpowered stage-specific analyses.
- Minor pre-processing alterations significantly impacted biomarker prognostic ability, changing validation success (P < 0.001 vs. P = 0.348).
Conclusions:
- Two NSCLC prognostic biomarkers were successfully validated.
- Validation success is unexpectedly sensitive to subtle data analysis decisions.
- A novel ensemble-based algorithmic approach was developed to enhance biomarker robustness and provide confidence metrics.
