Fig. 5. Two axes of covariance between biological mechanisms and symptom severity in PD.

a Based on a permutation analysis, two latent SVD components were significant or near-significant, explaining 48.4% (P = 0.001, FWE-corrected) and 13.2% (P = 0.069, FWE-corrected) of the covariance respectively. a, b High correlations of r = 0.70 (P = 3.11 × 10⁻¹¹) and 0.86 (P = 3.75 × 10⁻²¹), between the projections of statistically stable (based on 95% confidence intervals from bootstrapping) biological mechanisms and rates of clinical decline onto the latent components were observed. c, d Bootstrap ratios of each clinical assessment to the two latent components, providing a relative ranking of motor, nom-motor, psychiatric and cognitive domains. These saliences are proportional to the contribution of each term relative to every other term, for example showing that MDS-UPDRS scores, SDM and HVLT scores are the top contributors to the primary axis. Details about specific scores can be found in Methods: Clinical scores.