Abstract

This disclosure specifies the design of a privacy-preserving, multimodal longitudinal cohort study that links, for each consenting participant, five data layers: genome data (DNA sequence or genotyping array); date, time and place of birth, with the astronomical positions computed from them; a continuous record of prosocial transfers on a peer-to-peer gratitude ledger; continuous respiration data from a chest-worn wearable sensor; and verified kinship links in a pedigree graph, at a target of 100 million participants over decades. Consent is tiered: granted and withdrawn separately for each layer, and never tied to platform benefits. The privacy stack combines differentially private analysis under a published privacy budget, federated analysis with homomorphic encryption for genome data, on-device processing of respiration signals with differentially private uploads, per-jurisdiction data residency, and withdrawal by destruction of the participant's encryption key (crypto-shredding), which stops future use but cannot reach analyses already published. Governance comprises IRB-grade ethics review, pre-registered hypotheses, and publication of methods, code and aggregate findings, with individual-level data never released. Birth-time features are analysed as a coordinate label in association tests, not as a causal hypothesis; prior tests, including a comparison of 2,101 people born minutes apart, found no association. Population biobanks, genealogy-linked genotype resources and birth cohorts already combine several of these layers. Open questions are stated: the enrolment of minors, relatives whose genomes the data partly reveal, custody of the linkage between layers, the fate of the data if the institution ends, and oversight of an AI analyst.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.

Download

Share

COinS