Feasibility study of deep learning based radiosensitivity prediction model of National Cancer Institute-60 cell lines using gene expression

• Published in: Nuclear engineering and technology Vol. 54; no. 4; pp. 1439 - 1448
• Main Authors: Kim, Euidam, Chung, Yoonsun
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.04.2022; Elsevier; 한국원자력학회
• Subjects: Convolutional neural network; Deep learning; Gene expression; Prediction; Radiosensitivity; Survival fraction at 2Gy; 원자력공학; Deep learning; Radiosensitivity; Gene expression; Survival fraction at 2Gy; Convolutional neural network; Prediction
• ISSN: 1738-5733; 2234-358X
• DOI: 10.1016/j.net.2021.10.020

We investigated the feasibility of in vitro radiosensitivity prediction with gene expression using deep learning. A microarray gene expression of the National Cancer Institute-60 (NCI-60) panel was acquired from the Gene Expression Omnibus. The clonogenic surviving fractions at an absorbed dose of 2 Gy (SF2) from previous publications were used to measure in vitro radiosensitivity. The radiosensitivity prediction model was based on the convolutional neural network. The 6-fold cross-validation (CV) was applied to train and validate the model. Then, the leave-one-out cross-validation (LOOCV) was applied by using the large-errored samples as a validation set, to determine whether the error was from the high bias of the folded CV. The criteria for correct prediction were defined as an absolute error<0.01 or a relative error<10%. Of the 174 triplicated samples of NCI-60, 171 samples were correctly predicted with the folded CV. Through an additional LOOCV, one more sample was correctly predicted, representing a prediction accuracy of 98.85% (172 out of 174 samples). The average relative error and absolute errors of 172 correctly predicted samples were 1.351±1.875% and 0.00596±0.00638, respectively. We demonstrated the feasibility of a deep learning-based in vitro radiosensitivity prediction using gene expression.