Abstract:

Objective     To combine the kernel principal component analysis (KPCA) and the logistic regression model to propose a KPCA-based logistic regression model for nonlinear association analysis of complex disease gene mapping. Methods    For association study of case-control research design, the kernel principal component analysis (KPCA) was performed on single nucleotide polymorphisms (SNPs) of a candidate region to construct the logistic regression model with kernel principal components as independent variables, and then the PTPN22 and RNF186 gene regions of rheumatoid arthritis (RA) data from GAW16 were analyzed to illustrate the effectiveness and practicability of the KPCA-based logistic regression model. Results    Application to the PTPN22 and RNF186 gene regions indicated that the KPCAbased logistic regression model could detect regions which could be detected by a single-locus test (PTPN22), and identify significant regions which could not be identified by a single-locus test (RNF186). Conclusion    As an effective nonlinear association study method, the KPCAbased logistic regression model can identify more susceptible regions.

Key words: Kernel principal component analysis; Logistic regression; Complex disease gene mapping; Association study