Fig. 5: Individual re-identification using phenotypic and PRS nose profiles.

a Cumulative Match Characteristic (CMC) curve for the individual re-identification model using the PRS of the five nose traits, as detailed in Fig. 4 and Supplementary Note 6. The curve is based on 50 rounds of 10-fold cross-validation, highlighting CMC1%, CMC10%, and CMC50% in orange crosses and labelled in the top left. b Receiver operating characteristic (ROC) curve for the individual re-identification model using the PRS of the same five nose traits, based on 50 rounds of 10-fold cross-validation. The average AUC is labelled on the top left. c Distribution of AUC values across different prediction validation scenarios in two European cohorts, RS and TwinsUK. The density plot shows the distribution of AUC null values obtained through 10,000 replicates, where each replicate involved PRS constructed from a number and MAF-matched SNPs randomly selected across the genome. AUC values achieved by our PRS models under different validation scenarios were compared with the null distribution using arrowed lines.