Fig. 1: Illustration of the study design.

This figure presents a nested case-control study using the UKB dataset. The solid horizontal bars indicate the estimated duration from blood biomarker assessments to the clinical diagnosis of psychiatric disorders. They start from baseline visits and reach the dates of diagnosis, or vice versa. For the purpose of analysis, we aligned the patients based on the diagnosis dates. On the new timeline, the data collected at the baseline became the data collected at different time periods before and after the diagnosis. We used a color gradient from dark blue (representing zero years to diagnosis) to light blue (indicating ten years to diagnosis). For each identified case, we matched ~3 controls. For the controls, their corresponding duration was matched to that of the individuals with psychiatric disorders. Subsequently, we applied the locally estimated scatterplot smoothing (LOESS) technique to visualize the temporal variations in psychiatric assessments and blood-based biomarkers related to the timing of disease duration. To reveal the specific temporal patterns of blood-based markers that change during disease progression, we clustered them into different groups based on their temporal trends. Finally, we tested the differential associations between the identified blood-based biomarkers and brain structures within the case-control framework, using a multivariable linear regression model for statistical analysis.