1H NMR–based metabolic profiling of serum for the detection of pancreatic cancer.

Abstract

193 Background: Pancreatic cancer (PC) is one of the leading causes of cancer deaths with a 5-yr mortality rate of 95%, and the lack of a suitable early detection method contributes to its poor prognosis. Metabolomics, the analysis of the metabolic profiles in biological samples such as serum and urine is emerging as an important tool to complement other “omic” techniques. In an effort to identify potential biomarkers for PC, we analyzed serum from PC patients (pts) focusing on altered metabolic profiles using 1H nuclear magnetic resonance (NMR). Methods: The metabolite profiles from serum samples consisting of 55 PC pts and 32 healthy controls were analyzed using NMR combined with advanced supervised and unsupervised multivariate statistical methods such as partial least squares discriminant analysis (PLSDA) and principal component analysis (PCA). A number of metabolite markers selected based on p values and logistic regression rank the importance of each potential marker. Statistically significant metabolites between cancer and controls were used to build a prediction model. Results: Based on multivariate logistic regression analysis of 20 targeted metabolites, 10 metabolite markers were selected from the variable selection process and used to build a regression model with high accuracy (AUROC >0.99), a sensitivity of 95% and specificity of 95% using a training set of samples. When the model was tested on an independent set of patient samples, it yielded a sensitivity of 95% and a specificity of 100% (AUROC >0.98). Box and whisker plots for individual markers verified the high performance of all 10 markers. Conclusions: The high sensitivity of the metabolic profile that distinguishes PC pts from controls indicates the potential utility of 1H NMR metabolic profiling for the early detection of PC. The investigation has identified perturbations in several pathways such as glycolysis and amino acid metabolism, highlighting their contribution to disease onset. This study demonstrates the potential of metabolite profiling as an important tool toward detecting PC development. Future studies will involve metabolite validation on high risk pts, and additional mass spectrometry based metabolic discovery efforts. No significant financial relationships to disclose.