Centrifugal filtration of serum for FTIR spectroscopy does not improve stratification of brain tumours

Discrimination of brain cancerversusnon-cancer patients using serum-based attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy diagnostics was first developed by Handset al. Cameronet al. then went on to stratifying between specific brain tumour types: glioblastoma multiforme (GBM) vs. primary cerebral lymphoma. Expanding on these studies, 30 GBM, 30 lymphoma and 30 non-cancer patients were selected to investigate the influence on test performance by focusing on specific molecular weight regions of the patient serum. Membrane filters with molecular weight cut offs of 100 kDa, 50 kDa, 30 kDa, 10 kDa and 3 kDa were purchased in order to remove the most abundant high molecular weight components. Three groups were classified using both partial least squares-discriminate analysis (PLS-DA) and random forest (RF) machine learning algorithms; GBMversusnon-cancer, lymphomaversusnon-cancer and GBMversuslymphoma. For all groups, once the serum was filtered the sensitivity, specificity and overall balanced accuracies decreased. This illustrates that the high molecular weight components are required for discrimination between cancer and non-cancer as well as between tumour types. From a clinical application point of view, this is preferable as less sample preparation is required.