Metabolic Profiling of Prostate Tissue Using the K-Means Cluster Analysis

• Activity Number: 365
• Type: Contributed
• Date/Time: Wednesday, August 6, 2008 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Helena Gurascier*+ and Vickie Y. Zhang and Ying Lu and John Kurhanewicz
• Companies: University of California, San Francisco and University of California, San Francisco and University of California, San Francisco and University of California, San Francisco
• Keywords: k-Mean Cluster Analysis ; Discriminant Factor Analysis ; Metabolic Profiling ; Diagnosis ; Prostate Cancer

Abstract

Since PSA is not a specific cancer biomarker, prostate cancer will ONLY be detected in 25-40% of men biopsied. This paper investigates metabolic profiling of prostate tissue simultaneously using 16 metabolite concentrations. 108 TRUS guided biopsies were obtained from human prostates and frozen to -80o C. 1D spectra were acquired on these samples using high resolution magic angle spinning (HR-MAS) spectroscopy. The concentrations were quantified using the HR-QUEST algorithm. A k-means cluster analysis was performed followed by a discriminant factor analysis to obtain a 2-D plot. Three clusters were identified providing information on concentrations in the glandular, stromal and cancerous prostate tissue. These results indicate that healthy and cancerous prostate tissues have unique biochemical signatures that could be used to detect prostate cancer.