Construction of dose prediction model and identification of sensitive genes for space radiation based on single-sample networks under spaceflight conditions

To identify sensitive genes for space radiation, we integrated the transcriptomic samples of spaceflight mice from GeneLab and predicted the radiation doses absorbed by individuals in space. A single-sample network (SSN) for each individual sample was constructed. Then, using machine learning and genetic algorithms, we built the regression models to predict the absorbed dose equivalent based on the topological structure of SSNs. Moreover, we analyzed the SSNs from each tissue and compared the similarities and differences among them. Our model exhibited excellent performance with the following metrics: R2=0.980, MSE=6.74e−04, and the Pearson correlation coefficient of 0.990 (p value The topology structures of SSNs effectively predicted radiation doses under spaceflight conditions, and the SSNs revealed the gene regulatory patterns within the organisms under space stressors.