Abstract:

Objective    To explore the application of the geographically weighted regression(GWR) model in analyzing influencing factors of human population genetic structure. Methods    Using global ACE gene data and climate surveillance data, based on latent variable analysis and spatial statistical methods, synthesis factors from the climate variables were extracted by confirmatory factor analysis. Then spatial distribution of the ACE gene D allele frequency and the climate synthesis factors were estimated with the Kriging interpolation method. Finally the multiple linear regression model (global model) and GWR model (local model) were constructed to explore the relationship between the D allele frequency and the climate synthesis factors, respectively. Results     Two latent synthesis factors were extracted by confirmatory factor analysis. The multiple linear regression model showed that the two synthesis factors were both statistically related with D allele frequency (P<0.01). The local R2 and parameter estimation of each spatial unit of the GWR model displayed significant spatial variability, and its fitting effect was more desirable compared with the global model. Conclusion    The GWR model is more accurate in describing the spatial varying relationship between human population genetic structure and climate factors, and it is obviously superior to the global model.

Key words: Spatial genetics; Geographically weighted regression; Spatial autocorrelation; Spatial heterogeneity