论文部分内容阅读
Background: The identification and inference of path-specific effects along a selected subset of edges in complex network are very important in practical causal reasoning about legal, medical, public policy domains and system epidemiology,among others.We proposed a a riverway conflux-based non-parameter causal diagram model to identify the path-specific effect in system epidemiology.Results: Both simulation and application were conducted to evaluate its performances by comparing with methods including Permutation test and Bootstrap test (i.e.Standard Norm Bootstrap Confidence Interval, Basic Bootstrap Confidence Interval,Percentile Bootstrap Confidence Interval and Bias Correct Confidence Interval) in the scenarios of complex network graph.A serious of simulation studies showed that the Type Ⅰ error rate of above methods for estimating the total causal effect were stable when sample size increased to 1000.While the Type Ⅰ error rate of Permutation test for assessing path-specific effect showed stronger stability (i.e.close to the given nominal level 0.05) than Bootstrap test methods.Besides the power revealed increasing trend following the increase of the effect sizes of path-specific and the difference of path-specific effect between two groups.Furthermore, the path-specific effects were not influenced by the its parent nodes and child nodes.Application to real data of insulin signal pathway on the gene score of SNPs and DNA methylation data successfully identified the specific pathways (INS → INSR→ IRS → FOXO 1 → G6PC and INS → INSR → IRS → GRB2 → SOS) in metabolic syndrome (ms) group and non-ms data on the DNA methylation data and specific pathway (INS → INSR → IRS → AKT → PDE3) in the group of smoking group and non-smoking group on the gene score data.CONCLUSIONS:The proposed permutation-based PSEM are valid and powerful for detecting the specific pathway effect contributing to disease, thus potentially providing new insight and ways to unlock the black box of disease mechanism.