Abstract 2859: Deep learning recognizes FTD-resistant isolated cancer cells of colon cancer

Abstract

Abstract In recent years, innovative technologies that extract feature descriptions from the large volume of data on speech recognition, visual object recognition and detection as well as many other domains, such as drug discovery and DNA sequence annotations by deep learning techniques and applying them to automatic recognition etc. are drawing attention. As cancer research aiming at applying deep learning techniques to cases that are resistant to surgical therapy and drug therapy in metastatic colorectal cancer, we developed a fundamental technology that can predict the resistance of free cancer cells to fluorinated pyrimidine anticancer drugs by deep learning from the morphological image data taken from images. An experimental model was used in our investigation in order to clarify whether or not its image recognition ability can be applied to the determination of drug resistance of free cancer cells circulating in the peripheral blood. That is, a cell line established by inducing a resistance to FTD or 5 FU added to the cell culture solution was prepared over several months and the ability to recognize the tolerance of the drug was examined from a large volume of image data, and it was shown that it can be distinguished dominantly in a short-term culture system. Further, as a result of examination after separation at the single cell level, it was possible to distinguish fluorescent-labeled resistant strains dominantly. In addition, we were able to recognize the drug resistance character well by injecting resistant strains intravenously into the mice to prepare a model of free cancer cells and collecting circulating free cancer cells. Moreover, as a pre-clinical model, resistant strains were mixed with susceptible strains at various ratios and transplanted into mice and experimented. As a result, the nature of the resistance to treatment was predicted by image recognition, and death of the mice due to cancer was well correlated with the malignant trait of drug-resistant cancer cells. Then, by linking the feature expression obtained from the image and the Omics data, a detailed stratification of treatment resistance was possible. From the above, a technique in the mouse that can distinguish free cancer cells collected from the peripheral blood by deep learning of images was constructed, and a foundation to be applied to medical treatment and precision medical care in the future was established. Citation Format: Kiminori Yanagisawa, Masamitsu Konno, Masayasyu Toratani, Hirohiko Niioka, Ayumu Asai, Jun Koseki, Kenta Tsunekuni, Taroh Satoh, Kazuhiko Ogawa, Jun Miyake, Yuichiro Doki, Masaki Mori, Hideshi Ishii. Deep learning recognizes FTD-resistant isolated cancer cells of colon cancer [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 2859.