大量的研究表明微小RNA（microRNA, miRNA）是复杂疾病的重要生物标志物，miRNA导致复杂疾病发生，往往通过调控基因表达。因此，如何构建miRNA-mRNA调控网络和miRNA的效应分析，是本研究关注的要点。目前效应分析需要面对三个问题：（1）miRNA-mRNA多对多的调控关系复杂性、生物信息学预测假阳性高和缺乏疾病特异性；（2）多对多关系产生大量交互作用，传统的统计分析策略假阳性过高；（3）多因素中介模型分析缺乏有效分析策略。基于此，本课题拟结合靶标数据库和序列信息，基于多重混合效应模型，构建疾病特异性的miRNA-mRNA互作网络；基于混合效应模型和核函数，构建miRNA-mRNA及其交互作用的联合检验统计量，并推导其理论分布；将贝叶斯层次模型应用于中介分析模型，实现统计分析过程。本研究拟形成一整套效应分析的策略，可以有效克服目前的问题，具有重要的科学意义和应用价值。

Emerging studies have identified microRNA (miRNA) as biomarker to predict the complex diseases. However, miRNAs cause the diseases by regulating the expression of genes. Therefore, this proposal aims to constructing the regulatory network of miRNA-mRNA and analyzing the effect of miRNA, mRNA and their interactions. Several methods have been proposed, which also have some limitations, including: a) the prediction of bioinformatics has high false positive rate and is lack of specificity of diseases for the complexity of the relationships between miRNA-mRNA; b) traditional analysis is ineffective for the amount interactions; c) the analysis of multiple mediators model is difficult. Based on these considerations, this proposal focuses on three aspects. First, combing the target databases and sequence information, the relationships between miRNAs and mRNA is explained by multi-variate mixed model. Second, based on the similarity of mixed effect model and kernel machine regression, the omnibus test statistics is build, and its theoretical distribution is derived. Third, the Bayes hierarchical model is applied to solve the estimation and testing of mediation analysis. This study can overcome the drawbacks mentioned above, takes insight to the trancriptome data and has impact on the application of –omics data.