Fig. 5: Real-world application studies of the PHD platform.

a An example of using the PHD platform for data integration and visualization at an individual level; data collected through the iPOP study is displayed: a healthy event on 3/31 vs. viral infection event on 6/19. The PHD app provides an easily accessible interface (dashboard) for multiple data types, integration of surveys, and health-related predictions or information for real-time health monitoring for wearable, multi-omics, and clinical data. b–e An example of using the PHD platform for data analysis at a cohort level. b A UMAP⁴⁷ visualization of the cohort’s wearables data. Colors represent individuals and differences in individual wearables data are visible. Each dot represents 24 h of wearables data. c Demonstrates the same feature space as in (b), but labeled by the insulin status of the respective individual. Red color represents insulin resistant, green color represents insulin-sensitive individuals, respectively, while individuals of unknown status are indicated in gray. Separation by disease status is visible. For panels b and c, the diamond symbol represents the median of an individual and random IDs (e.g., 54nz) are annotated. d A boxplot summarizing the distribution of area under the receiver operating characteristics (AUROC) of the logistic regression models trained on user-level, aggregated features for the binary classification problems of IR vs IS (insulin-resistant vs. insulin-sensitive) and male sex in a 10× 5-fold cross-validation. e Shows the distribution of average precision scores of the classification. In both plots, the distribution consists of N = 10 measurements of performance of the outer cross-validation loop. The box indicates the interquartile range (IQR), with the central horizontal line indicating the median. The whiskers extend to the last non-outlier data-point within 1.5*IQR. f An example of applying the PHD platform for a COVID-19 detection study³⁰. Data from 2831 participants have been collected through the PHD platform. Summary of collected wearable data from various manufacturers, and clinical surveys are reported in the corresponding tables.