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

A method comprising:

• obtaining a urine sample from a subject;
• obtaining Raman spectra from the urine sample;
• comparing the Raman spectra of the urine sample to a selected model;
• wherein the selected model is constructed from various Raman spectra of urine from:
• one or more individuals having bladder cancer;
• one or more individuals not having bladder cancer;
• and
• one or more individuals having kidney disease, renal cancer, prostate cancer, and/or a non-neoplastic disease, and by:
• (a) applying baseline correction to a range of wavenumbers of the various Raman spectra to obtain baseline corrected Raman spectra;
• (b) performing normalization of the baseline corrected Raman spectra to obtain normalized Raman spectra;
• (c) performing principal component analysis (PCA) of the normalized Raman spectra to identify principal components (PCs) of the urine;
• (d) performing discriminant analysis of principal components (DAPC) to obtain one or more chemometric models based on one or more of the PCs, the one or more chemometric models comprising canonicals equal in number to the PCs;
• (e) for the DAPC analysis, determining a fractional contribution of at least four selected wavenumbers of the range of wavenumbers to each of the canonicals of the one or more chemometric models to determine which of the selected wavenumbers give rise to separations seen in a plot of two or more of the canonicals, wherein the selected wavenumbers correspond with at least nucleic acids, proteins, blood cells, and inflammatory cells;
• and
• (f) testing one or more of the chemometric models, using a leave-one-out or multi-fold cross-validation technique to select one of the chemometric models as the selected model having a desired level of accuracy, sensitivity, specificity, positive predictive value, and/or negative predictive value;
• wherein the selected model is one where the chemometric model is one comprising an accuracy of 100%, a sensitivity of 100% and/or a specificity of 100%;
• wherein the comparing of the Raman spectra of the urine sample to the selected model comprises classifying the urine sample according to the selected model as being urine from a subject who has bladder cancer, and the stage thereof;
• and
• administering cystoscopy to the subject whose urine sample has been classified according to the selected model as being urine from a subject who has bladder cancer and confirming the severity or stage thereof.