A Differential Privacy Preserving Framework with Nash Equilibrium in Genome-Wide Association studies
Ziwei Han, Hai Liu, Zhenqiang Wu · 2018
In genome-wide association studies (GWAS), inappropriate disclosure of genotype dataset may put individuals' privacy such as disease status, phenotype, kinship and individual identity at risk. The current privacy preserving mechanisms for GWAS mainly focus on protecting the statistical values or the query mechanisms and just achieve the tradeoff but not the equilibrium between utility and privacy. In this paper, we developed and evaluated our differential privacy framework and mechanism for raw genotype datasets in GWAS, and the framework finally achieves the Nash equilibrium between utility and privacy. Firstly, we code the raw genotype dataset to a genotype matrix based on the general genetic model in our mechanism. Then according to the expected utility and privacy, we add noises to raw matrix and the output matrix remains within the original range of p-value and its disturbance is as large as possible, which means the utility is available and the more deviation between raw dataset and output dataset, the larger privacy we get. Our framework can help release the disturbed dataset with raw format, which can promote the study of bioinformatics and biostatistics. It can also be used in a similar scenario such as privately publication of electronic medical record, or the user data which can be privately counted to a 2×2 contingency table.