Fig 5. Feature selection by the filter method after Index Construction in the VMR datasets.
(A–C) Volcano plots showing the features from the light-on dataset, matricized by the Total Distance, , and indices, respectively. (D–F) Volcano plots showing the features from the light-off dataset, matricized by the Total Distance, , and indices, respectively. A feature was deemed significant and selected from the matricized indices if the Holm adjusted p-value from the Wilcoxon rank sum test (in −log10(p); Y-axis) and the fold change (in ; X-axis) of that feature between the wildtype and Q344X mutants exceeded the cutoffs. The cutoff for p-value is −log10(p) > 2 (i.e., p < 0.01) for both the light-on and light-off VMR (horizontal solid lines), whereas the cutoffs for were and for the light-on and light-off VMR, respectively (vertical dashed lines). The selected features were highlighted with red dots when the wildtype’s values were significantly greater than those of the Q344X mutant. They were highlighted with black dots when the feature values between the wildtype and the Q344X were not significantly different. The feature label indicates the name of the index and the corresponding time point after the light change, separated by a dot (“.”). For example, the “” refers to the Total Distance in the first second after the light change.