A Powerful Statistical Method For Detecting Gene-Gene Interactions

Human diseases are mostly complex diseases caused by genetic and environmen-tal factors.At present,Genome Wide Association Study has successfully identified thousands of SNP loci associated with complex human diseases.These new findings have greatly promoted the development of medicine and biology.With the contin-uous in-depth researchs and explorations of researchers in various countries,GWAS has gradually become a hot issue in recent years.Recent years,in GWAS researches,researchers in various countries of the world have identified the association between thousands of SNPs and different traits or diseases,and provide information about the substance of the disease mechanism.Although a great progress has been made in the association test of genes and dis-eases,the identification of SNPs associated with diseases in GWAS only explains the percentage of the percentage of genetic variations,which leads to a question where and how the censored heredities can be identified.The possible explanation includes the following points:on the one hand,At present,the main effect is not enough to explain all the genetic effects,and the interaction effect has been confirmed in asso-ciation analysis.Based on this,a large number of test statistics have been proposed to test the interaction effect between genes and genes in recent years.On the other hand,the power of statistics widely used to test the associations between gene and gene or gene and environment is generally low.For example,in traditional logistic regression method,the interaction parameters are estimated to be inaccurate,and their test powers are relatively low.In this paper,under the case control experiment,we assume that the Hardy Weinberg equilibrium law(HWE)and the linkage equilibrium condition are satisfied in the control group,a new score test statistic is derived to improve the power of testing the interaction between gene and gene and we also construct the variance asymptotic estimation of the score test statistic by using the ? method.In this paper,we discuss the interaction between gene and gene in two different ways of genetic coding:1,the main effects and the interaction effect of G1 and G2 all adopt additive coding,2,the main effects of G1 and G2 adopt co-dominant coding while the interaction effect of G1 and G2 adopts additive coding.We have proved that under the two different genetic coding,the score test statistic and its asymptotic estimation of the variance are exactly the same.The final simulation shows that,under two different genetic coding,the score test statistic proposed by us has obvious numerical superiority compared with the conventional logistic method.