Detecting Early Warning Signal For Influenza A Disease

Influenza A,as an extremely contagious viral disease,exhibits a dynamic progression similar to a complex system,with the potential for rapid deterioration.However,relying solely on clinical symptoms to determine the critical threshold is challenging.Therefore,detecting early-warning signals is of great significance for the prevention and treatment of influenza A.Traditional methods for early-warning signals detection,such as the Dynamic Network Biomarker(DNB)approach,rely on changes in the correlation coefficients of gene expressions to reflect changes in system states.However,these methods are computationally complex and sensitive.To address these issues,this paper proposes two single-sample-based models for early-warning signal detection using H3N2 and H1N1 influenza A time-series gene expression datasets and KEGG signaling pathways:1)Single-sample Signal Perturbation model: Abnormal perturbations in signaling pathways can affect the expression of downstream genes,thereby disrupting normal biological functions and leading to disease occurrence.This model evaluates the perturbation level of genes by measuring the influence of their first-order neighbors in a specific signaling pathway,and then sums up the perturbation levels of all genes with weights to obtain a global warning indicator.The results show that the model can effectively detect early-warning signal for a single sample,with the advantages of simple calculation,earlier and more reliable warning time compared to traditional models based on correlation coefficient changes.2)Dysregulated Dynamic Network Biomarkers(dysregulated DNBs)model: The change in the expression of a single gene is merely a surface phenomenon of the system’s state,whereas the change in the interactions between genes accurately reflects the overall state of the system.This model selects abnormal regulations as dysregulated DNBs and calculates the dysregulation score of these biomarkers to obtain the warning indicator.Cross-validation results show that the model is simple and direct,and only requires the detection of four dysregulated DNBs to generate warning signals,making it suitable for clinical practice.In addition,the genes of dysregulated DNBs are significantly enriched in influenza-related pathways,indicating their biological significance.