DeepFlu: a deep learning approach for forecasting symptomatic influenza A infection based on pre-exposure gene expression

Abstract

Background and Objective: Not everyone gets sick after an exposure to influenza A viruses (IAV). Al-though KLRD1 has been identified as a potential biomarker for influenza susceptibility, it remains un-clear whether forecasting symptomatic flu infection based on pre-exposure host gene expression might be possible. Method: To examine this hypothesis, we developed DeepFlu using the state-of-the-art deep learning ap-proach on the human gene expression data infected with IAV subtype H1N1 or H3N2 viruses to forecast who would catch the flu prior to an exposure to IAV. Results: The results indicated that such forecast is possible and, in other words, gene expression could reflect the strength of host immunity. In the leave-one-person-out cross-validation, DeepFlu based on deep neural network outperformed the models using convolutional neural network, random forest, or support vector machine, achieving 70.0% accuracy, 0.787 AUROC, and 0.758 AUPR for H1N1 and 73.8% accuracy, 0.847 AUROC, and 0.901 AUPR for H3N2. In the external validation, DeepFlu also reached 71.4% accuracy, 0.700 AUROC, and 0.723 AUPR for H1N1 and 73.5% accuracy, 0.732 AUROC, and 0.749 AUPR for H3N2, surpassing the KLRD1 biomarker. In addition, DeepFlu which was trained only by pre-exposure data worked the best than by other time spans and mixed training data of H1N1 and H3N2 did not necessarily enhance prediction. DeepFlu is available at https://github.com/ntou-compbio/DeepFlu . Conclusions: DeepFlu is a prognostic tool that can moderately recognize individuals susceptible to the flu and may help prevent the spread of IAV.