Contribution of microarrays of gene expression (MAGE) to the definition of PET/CT as a qualified biomarker of early response in metastatic patients

11573 Background: Proliferating cancer cells consume elevated quantity of glucose, converted into lactate regardless the presence of oxygen (Warburg effect). This effect has been useful for imaging metabolically active tumors with FDG-PET, although its use in early response is controversial. Molecular mechanisms of FDG uptake are not fully understood. We have used MAGE to determine the most relevant genes involved in FDG uptake. Methods: Fresh-frozen tumor biopsies and quantitative basal FDG-PET/CT were obtained from metastatic lesions in cancer patients. Total tumor RNA was hybridized to a whole human genome oligonucleotide microarray. Gene expression signature-based prediction, using the most relevant genes involved in FDG uptake measured by SUV, was finally determined by Partial Least Squares (PLS). The interpretation of biological phenomena (IBP) derived from the selected genes was made by means of different public statistical bioinformatics resources. Results: 71 patients with different histological diagnosis were included in the training cohort and 13 in the validation one. 909 probes correlated significantly with SUV: 333 positively and 576 negatively. A predictive signature based on these 909 probes was built using PLS-3, with an RMSE in the validation set of 0.645 (within the 95% CI of RMSE determined in the training set). In IBP, other biological processes were more relevant than glycolysis in FDG uptake: RNA processing, ribosome biogenesis, protein processing, cell adhesion, cytoskeleton organization, angiogenesis and autophagy. Conclusions: This PLS-3-built signature is the first reported one that can accurately predict SUV. FDG uptake is a complex phenomenon that involves multiple biological processes, confirming the value of PET/CT in early response.