Abstract:
|
Traditional statistical methods for confidentiality and privacy protection of statistical databases do not scale well to deal with genome-wide association study (GWAS) databases, especially in terms of guarantees regarding protection from linkage to external information. The more recent concept of differential privacy is an approach that provides a rigorous definition of privacy with meaningful privacy guarantees in the presence of arbitrary external information.
A series of publications (Yu et al. (2014), Johnson and Shmatikov (2013), Uhler et al. (2013)) have explored differentially private methods for releasing single-nucleotide polymorphisms (SNPs) that are most relevant to a phenotype (e.g., a disease). In this talk, we will present new results on how to improve the aforementioned methods. In particular, we present a graphical interpretation of the chi-square statistic in a GWAS setting and an efficient method of computing the Hamming distance score, which is first proposed in Johnson and Shmatikov (2013) and shown in Yu et al. (2014) to have the best performance among the methods.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.