Table 1.
Parameter values and corresponding units for simulated datasets of synthetic nematodes.
For each parameter value a 10% variance is included to account for variability in nematode motility behaviour according to published data [30].
Fig 1.
Instantaneous snapshots of synthetically-generated nematode postures.
Sample postures shown for the range of (A) minimum and (B) maximum body amplitudes A as well as (C) minimum and (D) maximum body wavenumber k, according to Table 1. Corresponding supplementary videos are available in the SM.
Fig 2.
Phenotyping analysis for synthetic nematode data (range shown in Table 1).
Left column: distance matrix of average Euclidean distances between classes for parameter sweep in amplitude A (A), body frequency f (D) and body wavenumber k (G). Middle column (B, E, H): corresponding matrix for p-values obtained when comparing pairwise a given class with another, following a non-parametric two-sample test for multivariate samples using the minimum statistical energy test [35]. Note that matrices only show values in the upper-triangular region (due to symmetry) where the diagonal is not computed. Significance (black tiles) is set for a confidence level of 95% (p < 0.05). Right column (C, F, I): corresponding branching diagrams (i.e. dendrograms) that represent a hierarchy based on the relationships of similarity among different classes.
Fig 3.
Instantaneous frames of nematode crawling assays.
(A) Sequence showing a crawling nematode kept in the center of the field of view (FOV) using a motorized stage and custom software that automatically tracks the worm; data taken from [25]; see Results section for “Nematode motility data I”. (B) Sequence showing a nematode crawling in a fixed FOV; see Results section for “Nematode motility data II”. In both (A) and (B), time-lapse sequence shown for every 5th frame and nematodes are approximately 1 mm long.
Fig 4.
Locomotion similarity between 15 mutant strains.
(A) Distance matrix of average Euclidean distances between mutant strains (strains are ordered alphabetically). (B) Corresponding matrix for p-values for pairwise strain comparisons. Black tiles indicate significant differences, p < 0.05 using the minimum energy test. (C) Hierarchical clustering (dendrogram) of mutant strains based on the distance matrix in A. The coloured bars indicate strains that are expected to cluster together based on known functional relationships or, in the case of acd-5 and asic-2, previous clustering results from other methods [23].
Fig 5.
Locomotion similarity between 8 mutant strains.
(A) Distance matrix of average Euclidean distances between mutant strains (strains are ordered alphabetically). (B) Corresponding matrix for p-values for pairwise strain comparisons. Black indicates significant differences, p < 0.05 using the minimum energy test. (C) Hierarchical clustering (dendrogram) of mutant strains based on the distance matrix in A.
Fig 6.
Schematic flowchart of the SIFT-BoW algorithm.
See Methods section for details.