Power Analysis of Association Study Using Single Nucleotide Polymorphism

Jeong, Seong Hoon; Kim, Yong Sik

Abstract

In order to find the disease susceptibility gene in these complex genetic trait, there have been much interests in association study using single nucleotide polymorphism (SNP). Association study can be divided into two approaches candidate gene approach and linkage disequilibrium mapping. Recently, the candidate gene approach also has attracted much attention with the possiblity of whole genome wide scan being widely discussed. In genome wide scan, the amount of information that a locus can provide about the other adjacent loci becomes an important matter. When a locus was found not to be associated with a trait, it is questionable that closely situated adjacent loci can also be excluded as disease susceptible loci. To approach this problem theoretically, this study tried to find a method to calculate power in case-control association study, and aimed to investigate the influence of hypothesized inheritance model to the obtainable power. In addition, this study investigated the implication of negative association results in other adjacent loci. METHOD: The power of associatiation study was calculated applying non-centrality chi-distribution approximated by Poisson distribution. Using this method, the powers in each inheritance model were calculated in candidate gene approach. The power of chi-square test in linkage disequilibrium mapping was also algebraically obtained. This method was applied to simulation data to verify the validity of the method. RESULT: The proportion of phenocopy and the allele frequency of candidate locus exert substantial influence to the power of the study rather than penetrance matrix or inheritance model. The power in linkage disequilibrium mapping exponentially decreased according to the degree of linkage disequilibrium as anticipated, however, the marker allele frequency exert enormous influence to the power. Without any a prior knowledge about which marker allele had linked with disease susceptibility allele, the marker with equal distribution of each allele showed the highest power.
CONCLUSION
The implication of negative results obtained in association study can only be determined by power analysis. In linkage disequilibrium mapping, if favorable power had been obtained, exclusion analysis of specific gene segment could be attempted. Taking into account the probable inheritance model and the marker allele frequency in designing association study, more efficient and rigorous study can be possible.