Abstract 1582: The coupling of MDSCs with a computational neural network (NN) to detect solid tumors

The goal of this study was to create a non-invasive cancer detection assay that analyzes flow cytometry data in an objective method. By using an artificial neural network (NN), we were able to distinguish between cancer patients (CPs) and benign/healthy donors (HDs) based upon the flow cytometry profiles of MDSCs and other leukocytes. Myeloid-derived suppressor cells (MDSCs) are known to be key contributors in supporting tumor progression and tumor escape through their ability to suppress anti-tumor responses mediated through T cell and natural killer (NK) cell activity. Several studies have demonstrated their utility as indicators of tumor progression and possible predictors of clinical outcomes, but there is significant overlap with healthy individuals preventing discrete and accurate calls. We used standard multiparametric flow cytometry techniques to immunophenotype MDSCs and other leukocytes found in the peripheral blood of 65 biopsy-confirmed CPs with solid tumors and 84 HDs. A series of NNs utilizing pattern recognition computational algorithms are then created using three data sets: 1) the training set - this ‘teaches9 the two output categories of cancer and not cancer, 2) the validation set - this uses backpropagation to improve the accuracy of the trained networks, and 3) the testing set - this is used to rank the trained networks against each other. Finally, a naive testing set is then used to determine the overall sensitivity and specificity for the top-ranking networks. Using traditional flow cytometry gating methods to analyze MDSCs as a biomarker for cancer detection, it is difficult to achieve high sensitivity and specificity due to the substantial overlap with healthy individuals. Here, we incorporated a standard 12 marker flow cytometry assay with NN technology to achieve a sensitivity of 92% and a specificity of 89%. Pairing the advanced analytical capabilities of our NN with surface biomarker based analysis of MDSCs and certain leukocytes measured in peripheral blood has enabled us the ability to objectively identify patterns indicative for the existence of a solid tumor. Citation Format: George A. Dominguez, Kristen Maslar, Alexander Polo, Cyrus Sholevar, Anthony Campisi, John Roop, Dmitry Gabrilovich, Frank Rauscher, Amit Kumar. The coupling of MDSCs with a computational neural network (NN) to detect solid tumors [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 1582.