Assessing and managing risk when sharing aggregate genetic variant data

Abstract

Access to genetic data across studies is an important aspect of identifying new genetic associations through genome-wide association studies (GWASs). Meta-analysis across multiple GWASs with combined cohort sizes of tens of thousands of individuals often uncovers many more genome-wide associated loci than the original individual studies; this emphasizes the importance of tools and mechanisms for data sharing. However, even sharing summary-level data, such as allele frequencies, inherently carries some degree of privacy risk to study participants. Here we discuss mechanisms and resources for sharing data from GWASs, particularly focusing on approaches for assessing and quantifying the privacy risks to participants that result from the sharing of summary-level data.

Figure 1. Sharing 5,000 SNPs at different prevalence or prior probabilities

In the plots, we use simulations to show how the prior probability of being in a dataset impacts the ability to resolve if a person within a population using summary level allele frequencies from 5,000 SNPs on datasets of 500 individuals. In (a) we show a histogram of test-statistics based on the approach of Jacobs et al for resolving membership in 100,000 simulations when the person tested is actually within a dataset (red) and 100,000 simulations when the person tested is not within a dataset (blue). Since the simulations of being in a dataset and not within a dataset are equal, the prevalence or prior probability of being in the dataset is 0.5. In (b) we show 100,000 simulations when the person is not within the dataset (blue) and 100 simulations when they are within the dataset, equivalent to a prevalence or prior probability of being in the dataset of 0.001. The figures is zoomed to the right showing how a large number of tests of individuals not in the dataset can obscure the ability to distinguish true positive and false-positives. Describing risk as PPV allows one to consider prevalence for being in a dataset as a prior, thus increasing the accuracy in assessing the risk of a person within a dataset being correctly identified.