A simple pre-disease state prediction method based on variations of gene vector features

Background: The progression of disease can be divided into three states: normal, pre-disease, and disease. Since a pre-disease state is the tipping point of disease deterioration, accurately predicting pre-disease state may help to prevent the progression of disease and develop feasible treatment in time. Methods: In the perspective of gene regulatory network, the expression of a gene is regulated by its upstream genes, and then it also regulates that of its downstream genes. In this study, we define the expression value of these genes as a gene vector to depict its state in a specific sample. Then, we propose a novel pre-disease prediction method by such vector features. Results: The results of an influenza virus infection dataset show that our method can successfully predict the pre-disease state. Furthermore, the pre-disease state related genes predicted by our methods are highly associated with each other and enriched in influenza virus infection related pathways. In addition, our method is more time efficient in calculation than previous works. The code of our method is accessed at https://github.com/ZhenshenBao/sPGVF.git.