Methods of disease detection and characterization using computational analysis of urine raman spectra

Abstract

Disease detection and characterization using computational analysis of Raman spectra is used to detect disease-specific multi-molecular patterns 'spectral fingerprint' associated with specific diseases, cellular physiologic derangements, or altered metabolism from systemic reactions to disease. Comparison of the Raman spectral fingerprint of urine from subjects with specific diseases and those not (healthy persons) provides the means to identify key disease-associated changes in urine molecular composition. Methods include applying baseline correction to spectra of a desired wavenumber range e.g., with the Goldindec algorithm, or with ISREA and StaBAL; vector or specific band normalization; and one or more of principal component analysis (PCA); discriminant analysis of principal components (DAPC); principal least squares (PLS) regression, machine learning with neural networks (NN); identification of wavenumber loadings; calculation of total canonical distance (TCD); total spectral distance (TSD), total principal component distance (TPD); ANOVA; pairwise comparisons; and performing leave-one-out or multi-fold cross-validation analysis of chemometric models (DAPC, PLS, NN) to report predictive capabilities in terms of accuracy, sensitivity (true-positives), and specificity (true-negatives), positive predictive value (PPV) and negative predictive value (NPV).