Objective: To detect the gene-based gene-gene interaction by kernel canonical correlation analysis.
Methods: Based on case-control study, statistical simulation studies were conducted to construct and test the KCCU statistic to evaluate gene-based gene-gene interaction of gene FTO and gene PRDMl6 by kernel canonical correlation analysis.
Results: The power of KCCU statistic was related to significant level, sample size, minor allele frequency, and it was higher when the gene-gene interaction increased. The power arrived 0.8 at the significant level of 0.05 when the minor allele frequency was higher than 0.05, the interaction odds ratio was higher than 1.5, and the sample size was greater than 5000.
Conclusion: KCCU statistic is a valid and powerful statistical inference method for detecting gene-based gene-gene interaction in the large sample analysis with high interaction.