Computational protocol: Metabolome Profiling by HRMAS NMR Spectroscopy of Pheochromocytomas and Paragangliomas Detects SDH Deficiency: Clinical and Pathophysiological Implications12

Similar protocols

Protocol publication

[…] A combination of principal component analysis (PCA) and orthogonal partial least square discriminant analysis (OPLS-DA) was herein adopted. The PCA was performed to evaluate the quality of the data and to identify possible outliers . Then, the OPLS-DA was employed to optimize the separation between tumor subgroups. The following two-class models of OPLS-DA were built in this study:1.SDHx (n = 23) versus non-SDHx PHEOs/PGLs (n = 64) including sporadic, VHL, RET, NF1, and HIF2A tumors. This analysis was performed to confirm our previous results achieved in a smaller cohort of patients and obtained from the analysis of spectral intervals corresponding exclusively to four selected metabolites with a possible key role in tumoral development. Accordingly, succinate, glutamate, GSH, and ATP were identified and used in the model. Succinate amount was estimated by integrating the area comprised between 2.39 and 2.43 ppm, glutamate quantity between 2.32 and 2.38 ppm, GSH amount between 2.93 and 2.98 ppm, and ATP amount within the range of 6.07 to 6.11 ppm.2.SDHx (n = 23) versus sporadic PHEOs/PGLs (n = 48). Seventy-one selected tumor samples were included in this model to assess the global metabolic fingerprint of SDHx-related PHEOs/PGLs in an exploratory and untargeted manner. Accordingly, the full spectrum that ranged from 1 to 8.65 ppm was analyzed and considered in the statistical model. In this case, the OPLS-DA was performed on the whole set of metabolites (variables) to select those with a real discriminating power. Metabolites corresponding to variable importance for projection (VIP) value ≥ 1 were selected and labeled VIP metabolites. Two measurements of model quality were reported for OPLS-DA: R2Y and Q2 representing, respectively, the goodness of fit (i.e., data variation) and the goodness of prediction, as estimated by cross-validation. Q2 ≥ 0.5 can be considered as a good predictor .SDHx (n = 23) versus non-SDHx PHEOs/PGLs (n = 64) including sporadic, VHL, RET, NF1, and HIF2A tumors. This analysis was performed to confirm our previous results achieved in a smaller cohort of patients and obtained from the analysis of spectral intervals corresponding exclusively to four selected metabolites with a possible key role in tumoral development. Accordingly, succinate, glutamate, GSH, and ATP were identified and used in the model. Succinate amount was estimated by integrating the area comprised between 2.39 and 2.43 ppm, glutamate quantity between 2.32 and 2.38 ppm, GSH amount between 2.93 and 2.98 ppm, and ATP amount within the range of 6.07 to 6.11 ppm.SDHx (n = 23) versus sporadic PHEOs/PGLs (n = 48). Seventy-one selected tumor samples were included in this model to assess the global metabolic fingerprint of SDHx-related PHEOs/PGLs in an exploratory and untargeted manner. Accordingly, the full spectrum that ranged from 1 to 8.65 ppm was analyzed and considered in the statistical model. In this case, the OPLS-DA was performed on the whole set of metabolites (variables) to select those with a real discriminating power. Metabolites corresponding to variable importance for projection (VIP) value ≥ 1 were selected and labeled VIP metabolites. Two measurements of model quality were reported for OPLS-DA: R2Y and Q2 representing, respectively, the goodness of fit (i.e., data variation) and the goodness of prediction, as estimated by cross-validation. Q2 ≥ 0.5 can be considered as a good predictor .Cross-validation was used in each OPLS-DA model to determine the number of components and to avoid overfitting the data. A cross-validation embedded in a Monte Carlo resampling approach , was used during the construction of the model to build a confusion matrix that allowed a direct visualization of the performances of the model in terms of classification power [sensitivity (Se), specificity (Sp), positive predictive value (PPV), negative predictive value (NPV), and global accuracy]. The SIMCA P (version 11.0; Umetrics AB) package was used for statistical data analysis.Continuous variables are expressed as mean ± SD. Statistical analysis was performed using IBM SPSS Statistics version 20 (IBM SPSS Inc, Chicago, IL). The Spearman non-parametric test was performed to determine the correlation between metabolites. Comparisons of tumor metabolite concentrations between SDHx-related and sporadic tumors were performed using a Mann-Whitney U test. The receiver operating characteristic (ROC) curves were used to evaluate the clinical utility of metabolite quantification in the diagnosis of SDHx mutation. Areas under the ROC curve, sensitivity, and specificity were determined using MedCalc version 13.2.2 (MedCalc Software, Ostend, Belgium). According to the ROC curve, values exhibiting the best accuracies were chosen as the threshold for a screening test. Linear regression model was used to examine the association between metabolite levels and genetic status with adjustment for tumor location and total catecholamine levels as explanatory variables. For multiple testings, adjusted P values were calculated using the false discovery rate (FDR) procedure with the SAS PROC MULTTEST statement . PCA was performed to visualize and analyze correlations between metabolite concentrations and SDHx status. PCA was done using the package FactoMiner (Multivariate Exploratory Data Analysis and Data Mining with R). Values of P < .05 were considered to be statistically significant. […]

Pipeline specifications

Software tools SPSS, multtest, FactoMineR
Applications Miscellaneous, Gene expression microarray analysis
Diseases von Hippel-Lindau Disease, Neoplasms, Paraganglioma, Pheochromocytoma