Human microbiome privacy risks associated with summary statistics

Abstract

Recognizing that microbial community composition within the human microbiome is associated with the physiological state of the host has sparked a large number of human microbiome association studies (HMAS). With the increasing size of publicly available HMAS data, the privacy risk is also increasing because HMAS metadata could contain sensitive private information. I demonstrate that a simple test statistic based on the taxonomic profiles of an individual’s microbiome along with summary statistics of HMAS data can reveal the membership of the individual’s microbiome in an HMAS sample. In particular, species-level taxonomic data obtained from small-scale HMAS can be highly vulnerable to privacy risk. Minimal guidelines for HMAS data privacy are suggested, and an assessment of HMAS privacy risk using the simulation method proposed is recommended at the time of study design.