Fig. 1: Overview of ethomic fingerprinting of natural movement behavior of individuals with DMD and healthy controls.

a, Overview of our system, in which ethomic fingerprints were extracted from the digital twin (avatar) of a participant, which was created using the suit data of natural movement behavior. A supervised machine learning algorithm, GP regression, was then used to derive the digital biomarker from the ethomic fingerprints. b, Probability distribution of joint angles at three exemplary skeletal joints (from left to right: right elbow flexion, left hip joint flexion, left knee joint flexion) for natural movement behavior data of individuals with DMD (blue) and healthy controls (red). Shaded regions denote the standard error and the stars denote a significant difference (P < 0.05, Kruskal–Wallis one-way ANOVA). c, Holistic view of whole-body coordination in ADLs. The correlation matrix (that is, the Pearson correlation coefficient of the joint angular velocities) for major skeletal joints is shown for healthy controls (left) and individuals with DMD (middle); the difference is shown on the right. The value in the color bars corresponds to the Pearson correlation coefficient. For a complete list of the joints for which movement data were obtained, see Supplementary Table 3.