Related Experiment Video
Updated: Jun 14, 2026
07:58
Cell-Free DNA Integrity Analysis in Urine Samples
Published on: January 5, 2017
13.9K
Non-invasive bladder cancer detection: Identification of a urinary volatile biomarker panel using GC-MS metabolomics
 Carapito1, V S Fernandes Ferreira2, A C Silva Ferreira3
1Associate Laboratory i4HB - Institute for Health and Bioeconomy, University of Porto, Porto, Portugal; UCIBIO - Applied Molecular Biosciences Unit, Laboratory of Toxicology, Faculty of Pharmacy, University of Porto, Porto, Portugal.
Talanta
|September 4, 2025
Summary
Urinary volatile organic compounds (VOCs) analyzed with machine learning show promise for non-invasive bladder cancer (BC) detection. An eight-VOC panel achieved 91% accuracy, outperforming current tests.
Area of Science:
- Biomarker Discovery
- Analytical Chemistry
- Computational Biology
Background:
- Early bladder cancer (BC) detection is challenging due to limitations of current diagnostic methods.
- Invasive, expensive, and insufficiently sensitive tests hinder early diagnosis.
- Metabolomics and machine learning (ML) offer potential for novel, non-invasive biomarkers.
Purpose of the Study:
- To identify urinary volatile organic compound (VOC) biomarkers for non-invasive bladder cancer (BC) detection.
- To evaluate the performance of machine learning algorithms in classifying BC based on VOC profiles.
- To develop a robust diagnostic panel for early BC detection.
Main Methods:
- Urinary VOCs analyzed from 87 BC patients and 90 controls using headspace solid-phase microextraction coupled with gas chromatography-mass spectrometry (HS-SPME/GC-MS).
- Twenty-seven identified VOCs used to train five ML algorithms: random forest (RF), support vector machine (SVM), partial least squares-discriminant analysis (PLS-DA), extreme gradient boosting (XGBoost), and k-nearest neighbors (k-NN).
- Model performance assessed via cross-validation and an independent validation set using AUC, sensitivity, specificity, and accuracy.
Main Results:
- Random Forest (RF) model achieved high performance (AUC = 0.913) with 27 VOCs.
- A selected eight-VOC panel demonstrated improved performance on the validation set (AUC = 0.872; accuracy = 91%).
- The eight-VOC panel (ketones, aldehydes, fatty alcohol, phenol) outperformed FDA-approved urinary assays and matched urine cytology specificity.
Conclusions:
- Urinary VOCs combined with ML show significant potential for non-invasive bladder cancer (BC) detection.
- The developed eight-VOC panel offers a promising, accurate, and sensitive diagnostic tool.
- Further large-scale validation is crucial for clinical translation and widespread adoption.
