Statistical Methodology

statistical methodology

Our research also involves exploring different statistical methods to treat DESI-MS imaging data.The application of multivariate techniques allows the entire chemical and spatial information contained in the DESI-MS images to be considered, which can guide a complete exploration and interpretation of the imaging data. Many different statistical approaches have been investigated, including PCA, PLS-DA and support vector machine. In cancer research, when using PLS-DA for classification of cancer and normal tissue, misclassification rates of 8% for clear cell renal cell carcinoma, 14% for papillary renal cell carcinoma and 12% for bladder carcinoma were achieved; encouragingly low values considering the small number of samples used so far. We are currently exploring different methods for data processing and novel statistical tools for building classifiers.