Identification of volatile signatures for non-invasive cancer detection using Secondary Electrospray Ionization (SESI) – High Resolution Mass Spectrometry and machine learning- based data analysis.

Identification of volatile signatures for non-invasive cancer detection using Secondary Electrospray Ionization (SESI) – High Resolution Mass Spectrometry and machine learning- based data analysis.

Miceli R (1), Locati L (2), Segrado F (3), Patricola P (3), Garrone G (3), Cavalleri A (3), Mancinelli M (2), Granata R (2), Agresti R (4), Brambilla P (5), Milani M (6), Licitra L (2,7) and Orlandi R (3)  

(1) Clinical Epidemiology and Trial Organization, (2) Head and Neck Medical Oncology Department, (3) Research Department, (4) Breast Surgery, Fondazione IRCCS Istituto Nazionale dei Tumori (INT), Milan, Italy (5) Department of Laboratory Medicine, University of Milano-Bicocca, Hospital of Desio, Desio-Milan, Italy   (6) UO Multidisciplinare di Patologia Mammaria e Ricerca Traslazionale, ASST Cremona, Italy (7) University of Milan, Italy

Full Poster:

http://ow.ly/8Xvk30mo8AR

Abstract:
Prospective studies were designed to identify cancer-related volatile organic compounds (VOCs) and develop non-invasive diagnostic tools for clinical practice in head and neck cancer (HNC) and breast cancer (BC). The first cohort includes 150 HNC patients and 100 controls (patient partners), and will be used to identify HNC-related and human papilloma virus (HPV)-related volatile signatures. An additional series of 100 controls will be analyzed for identification of gender-related, smoke-related and food-related VOCs profiles and environmental contaminants. A third cohort, including 200 early stage primary BC patients, 100 benign disease patients, and 200 controls with negative breast imaging will be used to identify BC-related and benign-related volatile signatures. Such signatures will be validated in a multicenter cohort of 400 BC patients and 200 controls. Exhaled breath is collected in sterilized nalophan bags and analyzed within two hours with SuperSESI (Fossiliontech) coupled with LTQ Orbitrap Elite (Thermo Fisher). VOCs detection occurs without any sample pre-treatment and chromatographic step in the 80–450 m/z range. Adherence to rigid Standard Operating Procedures in breath collection controls the pre-analytical and analytical variability. Data analysis integrates pre-processing for quality control, and a class prediction based on machine-learning techniques, including a robust feature selection, and a classifier development with internal validation. Our studies supports the value of real time detection techniques for clinical applications of breath analysis, and underscores the importance of sample quality assessment and quality control of data from breath analysis using a robust data pre-processing techniques to address unbiased pattern discovery.  This study is supported by Direzione Scientifica INT (Fondo per la Ricerca Istituzionale) and Ministero della Salute.