Gas Chromatography-Ion Mobility Spectrometry data analysis of urine samples to detect colorectal cancer.

The metabolic investigation of human populations is increasingly important in the study of health and disease. Urinary metabolic profiling plays a significant role in systems biology research, expanding its applications in clinical diagnostics, mechanistic studies, personalized health care, and molecular epidemiology. This study underscores the importance of analyzing volatile organic compounds (VOCs) using advanced characterization instruments, particularly for diagnosing colorectal cancer. Early detection through screening is crucial for effective treatment and improved prognosis in colorectal cancer, a prevalent malignancy characterized by abnormal cell growth in the colon or rectum, often detectable through biomarkers in urine. An in-depth quantitative study was conducted using a cohort of urine samples from colorectal cancer patients and non-cancer controls, employing Gas Chromatography-Ion Mobility Spectrometry (GC-IMS). This technique, known for its rapid, reliable, and cost-effective analysis of volatile mixtures, faces challenges due to the high dimensionality and complexity of its data, including baseline issues, misalignments, long peak tails, and strong nonlinearities that require extensive correction. A comprehensive data processing workflow using the GCIMS R package was applied to address these challenges. The dataset was analyzed using two feature extraction approaches: the full Reverse Ion Chromatogram (RIC) response and the peak table method. The capacity to deduce chemical information from the samples was evaluated by comparing the classification results obtained through Partial Least Squares Discriminant Analysis (PLSDA) and the Variable Importance for Projection (VIP) scores. Choosing a feature extraction strategy involves balancing the user's objectives, computational effort, and the desired depth of chemical information. However, analyte identification remains challenging in GC-IMS due to the limited resolution and size of electrical mobility databases, as well as the lack of comprehensive libraries for compound identification, which hinders accurate and efficient analysis. The present thesis highlights the potential of the GCIMS R package and peak table method in creating a reliable screening tool for colorectal cancer, potentially reducing the need for invasive procedures like colonoscopies. The findings lay the groundwork for further development of non-invasive diagnostic tools and emphasize the importance of robust data processing and feature extraction methods in metabolomics research.