Plasma nanoDSF Denaturation Profile at Baseline Is Predictive of Glioblastoma EGFR Status

SIMPLE SUMMARY: The toolkit for diagnosing the most aggressive primary brain tumor glioblastoma (GBM) is very limited. We recently demonstrated that plasma denaturation profiles (PDPs) of GBM patients and healthy controls obtained with nanoDSF can be automatically classified using artificial intelligence (AI) algorithms. Since PDPs have been shown to be useful for subtype differentiation for lung cancer, we decided to investigate whether nanoDSF-derived PDPs could also be used to discriminate EGFR alterations in GBM, which is important for determining therapy strategies. We found that AI is able to discriminate EGFR alterations in GBM with an 81.5% accuracy. Thus, we have demonstrated that the use of plasma denaturation profiles could answer the unsatisfied neuro-oncology need for a predictive diagnostic biomarker, which could complete MRI and clinical data, allowing for a rapid orientation of patients for a definitive pathological diagnosis and treatment. ABSTRACT: Glioblastoma (GBM) is the most frequent and aggressive primary brain tumor in adults. Recently, we demonstrated that plasma denaturation profiles of glioblastoma patients obtained using Differential Scanning Fluorimetry can be automatically distinguished from healthy controls with the help of Artificial Intelligence (AI). Here, we used a set of machine-learning algorithms to automatically classify plasma denaturation profiles of glioblastoma patients according to their EGFR status. We found that Adaboost AI is able to discriminate EGFR alterations in GBM with an 81.5% accuracy. Our study shows that the use of these plasma denaturation profiles could answer the unmet neuro-oncology need for diagnostic predictive biomarker in combination with brain MRI and clinical data, in order to allow for a rapid orientation of patients for a definitive pathological diagnosis and then treatment. We complete this study by showing that discriminating another mutation, MGMT, seems harder, and that post-surgery monitoring using our approach is not conclusive in the 48 h that follow the surgery.