1140 A Novel Diagnostic Prescreen to Detect Potential Intracranial Aneurysm From Circulating Blood MRNA

INTRODUCTION: Recently our research group along with Neurovascular Diagnostics, Inc. created a PCR-based intracranial aneurysm (IA) diagnostic, called AneuScreenTM, which analyzes a 50-gene panel in whole blood to predict lesion presence. METHODS: We collected blood samples from multiple centers across the USA to create a large cohort of 550 cases with and without unruptured IA. Prediction models, based on panel gene expression (from our previous published work) and basic patient information, were developed using machine learning algorithms. An assay was created in collaboration with DxTerity Diagnostics, Inc. Gene expression intensity (robustness), variability (variation), and linearity with input were tested during assay development. Training and testing of a support vector machine model were performed using 85% of the cases (split 80:20 for training and testing). Independent validation of model performance was done in the remaining 15%. RESULTS: During assay development, all 50 panel genes signaled within acceptable and detectable range (>100 RFU), had low coefficient of variation (<0.3), and acceptable linearity. We determined 50uL is the minimal required input volume for stabilized whole blood. During training, 12 features were identified as most important, and the final model performed well in the testing dataset without showing evidence of overfitting (AUC=0.74). Further, in the independent validation dataset of n=83 cases, the model reached a sensitivity of 0.84 and an AUC of 0.80. CONCLUSIONS: We developed an IA prescreen that performed with an AUC=0.80 and sensitivity=0.84.