Fig. 4.

Correlation and classification analyses between protrusion velocity and actin regulator dynamics. a–d Time-lag correlation analysis based on Pearson’s cross-correlation of edge velocity and actin (a), Arp3 (b), VASP (c), and HaloTag (d). Solid lines indicate population averages. Shaded error bands indicate 95% confidence intervals of the mean computed by bootstrap sampling. The number of samples used for the analysis is identical from Fig.3b–e. e Comparison and statistical testing of maximum correlation coefficients from a–d in each cluster. The error bar indicates 95% confidence interval of the mean by bootstrapping. **p < 0.01, ***p < 0.001 and ****p < 0.0001 indicate the statistical significance by two-tailed two-sample Kolmogorov–Smirnov (KS) test. The p-values are listed in Supplementary Table 3. f–h Time-specific correlation analysis based on pairwise Pearson’s correlation coefficients of protrusion velocity and fluorescence intensity time series registered relative to protrusion onset. The regions surrounded by the black lines are statistically significant correlation by Benjamini-Hochberg multiple hypothesis testing. i Pearson’s correlation coefficients between early Arp3 intensities and late protrusion velocities in each cluster. The error bar indicates 95% confidence interval of the mean by bootstrapping. The numbers of samples in this analysis are 204 (Cluster I), 112 (Cluster II-1), 161 (Cluster II-2), 178 (Cluster II-3) and 102 (Cluster III) respectively. j–k Classification analysis of Cluster III against Clusters I/II based on fluorescent intensity time series. Boxplots of the accuracy (j) and Matthews correlation coefficients (k) represent multiple classification results. RF stands for Random Forest, DNN for Deep Neural Network, and SVM for Support Vector Machine. The central line indicates median, and both edges of the box each represent 25^th and 75^th percentiles. The numbers of samples used in these analyses are 934 (actin), 757 (Arp3) and 682 (VASP) respectively