Johanna Olschewski: Clustering-Based Analysis of Cockroach Activity Rhythms
Abstract:
This thesis presents a clustering-based analysis of recovery interval patterns in male Madeira cockroaches (Rhyparobia maderae), with the aim of investigating potential circadian and ultradian rhythmic structures. We analyzed wheel-running activity of 28 cockroaches recorded under constant darkness for 40–95 days to reveal endogenous rhythm patterns independent of environmental cues.
By using Lomb–Scargle and Chi-square periodograms, we detected significant circadian rhythms in 26 of 28 cockroaches, with periods ranging from 22.95 to 24.66 hours (mean: 23.57 ± 0.40 hours). To investigate individual differences, we applied K-Means clustering and Self-Organizing Maps to recovery intervals of each individual cockroach. This revealed two behavioural groups. The "Olympic" group is characterized by mostly short intervals (0–3.3 hours), while the "Non-Olympic" group, in comparison, showed significantly more medium-length intervals (5.3–9.2 hours).
Further K-Means clustering of the recovery intervals, using a predefined number of K = 7 clusters based on the explained variance, revealed seven characteristic recovery intervals at 0.65 h, 1.73 h, 3.17 h, 5.31 h, 9.19 h, 16.48 h, and 23.60 h across the population, with the shortest three rhythms accounting for over 90% of all recovery periods. Bootstrap analysis and an additional analysis with a similar data set confirmed a high centroid stability for the four shortest rhythms.
This thesis demonstrates how unsupervised clustering techniques can reveal characteristic recovery patterns in biological data, providing insights into individual behavioural variations without assuming predefined rhythmic waveforms like traditional analysis methods. These methods may also provide a basis for further investigation of potential ultradian rhythms.
Zoom link for online participation:
https://uni-kassel.zoom-x.de/j/66183030603?pwd=3TYcaqxJjpBNz1uvOXMUfdpH0Q42r7.1
Meeting ID: 661 8303 0603
Passcode: 341157
We would be grateful if you could circulate this announcement within your departments.
Sincerely,
E. Friedmann