Advanced search
    Wang Zijian, Ye Meixia, Zhang Han, Wu Rongling. Mixed-effect model development for functional mapping[J]. Journal of Beijing Forestry University, 2024, 46(5): 163-172. DOI: 10.12171/j.1000-1522.20220416
    Citation: Wang Zijian, Ye Meixia, Zhang Han, Wu Rongling. Mixed-effect model development for functional mapping[J]. Journal of Beijing Forestry University, 2024, 46(5): 163-172. DOI: 10.12171/j.1000-1522.20220416

    Mixed-effect model development for functional mapping

    • Objective Using the abundance of Escherichia coli strains and functional mapping model as the research foundation, this study explored the impact of mixed effects on the performance of functional mapping model by introducing fixed effects of the population and random effects caused by kinship relationship among individuals into the functional mapping model.
      Method Based on the framework of functional mapping, this study employed the growth data from dynamic cultures of Escherichia coli as a practical case. Subgroups and SNP genotypes were considered as sources of fixed effects, and these fixed effect factors were integrated into the mapping model, leading to the extension of Q-matrix model. While maintaining the use of variance-covariance model for modeling random residuals, the Legendre model was employed to model random effects. A mixed-effect model analysis combining fixed effects with general random effects (model 1) was conducted. Additionally, the restricted maximum likelihood estimation method was utilized to derive variance-covariance parameters, random effects, and fixed effects, enabling the analysis of a mixed model combining fixed effects with random effects arising from kinship relationships (model 2). Finally, the Zwald test method was utilized to derive the calculation method for p-values at each marker locus.
      Result (1) In both models, 95% of the markers exhibited p-values that were consistent with the expected values, resulting in satisfactory upward curvature in the QQ plot. (2) Compared with model 1, model 2 detected a greater number of SNP loci, indicating that model 2 provided a stronger explanation for the random effects caused by kinship relationship. (3) Computer simulation results revealed that when the sample size was small and the heritability was low, the false-positive rate of the model was 4.77%. However, when the sample size reached 800 and the heritability was 1%, the discovery rate of quantitative trait loci (QTL) by the model can exceed 70%. Alternatively, when the sample size was 400 and the heritability exceeded 1.5%, the QTL discovery rate can also exceed 70%.
      Conclusion The mixed model approach proposed in this study, which introduces fixed effects and random effects caused by kinship relationship into the functional mapping model, effectively improves the theory of functional localization. This approach exhibits excellent calibration capabilities for covariate factors in fixed effects and can effectively dissect random effects from remaining residuals. This lays a solid foundation for subsequent improvements in functional localization and the development of software packages for the fixed effects plus kinship (Q + K) model.
    • loading

    Catalog

      Turn off MathJax
      Article Contents

      /

      DownLoad:  Full-Size Img  PowerPoint
      Return
      Return