Fig. 4: Longitudinal predictions of the clinical scales.

a, Scatter plot of the 6MWD from the visit at time T against the 6MWD from the visit at time T + 6 months from both the KineDMD study (n = 13 individuals with 24 longitudinal visits) and the Gemelli study (n = 44 individuals with 122 longitudinal visits). A small jitter was added to the points to show any overlapping points. b, Comparison of the aggregate RMSE of the LOSO cross-validated predictions of the 6MWD at T + 6 months by the ethomic fingerprints from visit T (blue bar) and the 6MWD at visit T (red bar). The LOSO cross-validated prediction of the 6MWD at T + 6 months using the 6MWD at T months was performed on the combined cohort of KineDMD (n = 13 individuals, 24 visits) and Gemelli studies (n = 44 individuals, 122 visits). The LOSO cross-validated prediction of the 6MWD at T + 6 months using the ethomic fingerprints at T months was performed on the KineDMD cohort (n = 13 individuals, 24 visits). c, Plot of the aggregate LOSO cross-validated RMSE of the predictions of the 6MWD at T + 6 months as a function of the number of individuals used to build the machine learning model. The model using ethomic fingerprints achieves better performance with a smaller number of individuals (marked by the vertical dotted lines) compared to the model using 6MWD from the combined data of the KineDMD and Gemelli studies. The shaded regions indicate the s.d. of the results from different models built for each number of individuals. d–f, Corresponding results for NSAA. g–i, Corresponding results for PUL.