Fig 1.
Diagram representing the interplay between, from left to right, the raw data, the rethomics packages (in blue) and the rest of the R ecosystem.
Fig 2.
Illustration of a behavr object, the core data structure in rethomics. The metadata holds a single row for each of the n individuals. Its columns, the p metavariables, are one of two kinds: either required—and defined by the acquisition platform (i.e. used to fetch the data)—or user-defined (i.e. arbitrary). In the data, each row is a ‘read’ (i.e. information about one individual at one time-point). It is formed of q variables and is expected to have a very large number of reads, k, for each individual i ∈ [1, n]. Data and metadata are implicitly joined on the id field. Note that the names used for variables and metavariable in this example are only plausible cases which will likely differ in practice.
Fig 3.
Tile plot showing the fraction of time spent moving as a colour intensity. Each individual is represented by a row and time, on the x-axis, is binned in 30 minutes consecutive epochs. A: Uncurated raw data. B: Data after the curation step. Time was trimmed and data from dead animals removed. The red ‘+’ symbols show animals that were removed from the subsequent analysis as they did not survive five complete days in LL. The white and black rectangles on the time axis show L and D phases, respectively. In the LL regime (for t > 0), grey rectangles represent subjective nights.
Fig 4.
Visualisation of the periodicity in the activity of eight selected animals.
A: Double-plotted actograms showing all activity during experiment. Time is defined relative to the transition from LD to LL (at day 0). B: χ2 periodograms over the LL part of the experiment matching the animals in A. The blue cross represents the highest peak (if present) above the significance threshold, at α = 0.05, (red line). Titles on top of each facet refer to the label allocated to each individual. See S2 Fig for all 53 animals.
Fig 5.
Population statistics on circadian phenotype.
A: Average periodograms. The aggregated relative power of the periodogram of all animals. The solid lines and the shaded areas show population means and their 95% bootstrap confidence interval, respectively. B: Frequencies of rhythmic animals. Number of rhythmic animals (i.e. with a significant peak) in each genotypes. Dark and clear fillings indicate rhythmic and arhythmic animals, respectively. C: Peak periodicity power and average. Values of the peak period for animals with a peak above the significance threshold, at α = 0.05(i.e. rhythmic). Individual animals are shown by dots whose size represent the relative power of the peak period. The error bars are 95% bootstrap confidence interval on the population mean.
Fig 6.
Wavelet analysis of positional data.
A: Raw position data for a representative female (top) and male (bottom) Drosophila melanogaster over five days, in black. The thick coloured lines show the average position every two hours. The green rectangles in the background show the two time windows selected for B and C. B: Close up of the first time window in A, showing position over one hour, starting at the beginning of the L phase of day 1. C: Close up of the second time window in A, showing position over one hour, starting at the middle of the L phase of day 1. D: Continuous wavelet transform spectrogram for the two representative animals. E: Average spectrogram across 40 males and 40 females. In D and E, the lines on the right show the marginal power spectra corresponding to the spectrograms shown—that is the average across 12 hours in either the light (white bar) or dark (black bar) phase. The male data was collected and described in our previous study [21] (controls in Fig 5M–5P) and the females data was acquired in parallel, in the same experimental conditions, but not previously published.