The Genetic Basis of the Human–Cannabis Relationship Revisited: Archival Recovery of a Deleted openSNP Cohort and a Permutation-Corrected Re-Analysis of Its 45-SNP Panel

In 2018 a preprint reported 45 SNPs associated with self-reported “affinity to cannabis” in 26 openSNP participants, including an outlier in MYO16, and proposed a qPCR panel to predict individual cannabis response. The platform that hosted those data, openSNP, deleted its entire holdings in April 2025. We show that the cohort is nonetheless recoverable, and we use it to test the original claims. From Wayback Machine captures of openSNP phenotype 401, the 2017 Internet Archive data dump and the 2020 Zenodo data-freeze of 5,393 genomes, we reconstructed 78 respondents, classified 54 by verbatim self-description, and recovered genotypes for 45 (31 positive, 14 negative affinity) — a 73% larger analytic sample than the original. Genotype decoding was validated by three independent checks. In a genome-wide screen of 69,530 QC-passed SNPs under a 1,000-permutation max-statistic test, no variant was significant (best SNP rs6009352, z = −2.48, empirical p = 0.19; zero SNPs at permutation p ≤ 0.05 at either of two QC thresholds). The MYO16 region showed no signal, and principal component 1 did not separate the phenotype groups, contradicting the original figure. The only nominal signal was CNR1 rs1049353 (z = −2.21, uncorrected p ≈ 0.027; Firth p = 0.071), with the T allele enriched among negative responders — directionally concordant with independent 2025–2026 adverse-event studies. Simulation shows that at this sample size 80% power requires a per-allele odds ratio of about 3.8 even at a nominal threshold, whereas every replicated cannabis-use locus has an odds ratio between 1.03 and 1.2. We conclude that the original panel was statistical noise arising from uncorrected thresholds and a circular selection-then-visualisation procedure, that single-marker prediction of cannabis response is unattainable at consumer-genomics scale, and that the defensible successor programme is pharmacogene-informed safety screening. We provide a fully documented archival reconstruction protocol.