A Novel EEMD-Based Privacy Preserving Approach for Top-k SNPs Query in Genome-Wide Association Studies

Genome-wide association studies (GWAS) have been a popular method for querying Top-k most significant sets of singe-nucleotide polymorphism locations (SNPs) to discover the genetic factors of diseases. Doctors have been increasing interest to in querying SNPs that significantly associated to diseases. However, the queries and GWAS data share service bring great privacy breach risk to the patients. In traditional Laplace privacy preserving mechanism, privacy budget ε is different to be determined, the Laplace noise is added to original data only once, which causes randomness and potential security risk to privacy protection. A novel approach is proposed to implement the differential privacy preserving for Top-k SNPs query in GWAS. Ensemble Empirical Mode Decomposition (EEMD) is a noise-assisted data analysis method, it is firstly proposed to be used as noise-adding approach in privacy protection mechanism. In the EEMD approach, random white noise will be added into the SNPs querying results for many times. Moreover, the parameters of added white noise can be precisely controlled to achieve better effect. PriSTRAT software tool was redeveloped to evaluate EEMD-based privacy preserving approach, the dataset was generated by PLINK tools. Experimental results show that EEMD-based privacy preserving approach has achieved higher accuracy than Laplace mechanism under the same noise-adding strength. The novel approach could be used to realize differential privacy preserving for Top-k SNPs query in GWAS to achieve the balance between privacy preserving and query accuracy.