Statistical analysis and dataset for: Acute exposure to caffeine improves foraging in an invasive ant

Linked to the journal article published in iScience (https://doi.org/10.1016/j.isci.2024.109935).

Abstract

Argentine ants, Linepithema humile, are a particularly concerning invasive species. Control efforts often fall short likely due to a lack of sustained bait consumption. Using neuroactives, such as caffeine, to improve ant learning and navigation could increase recruitment and consumption of toxic baits. Here, we exposed L. humile to a range of caffeine concentrations and a complex ecologically relevant task: an open landscape foraging experiment. Without caffeine, we found no effect of consecutive foraging visits on the time the ants take to reach a reward, suggesting a failure to learn the reward’s location. However, under low to intermediate caffeine concentrations ants were 38% faster with each consecutive visit, implying that caffeine boosts learning. Interestingly, such improvements were lost at high doses. In contrast, caffeine had no impact on the ants’ homing behavior. Adding moderate levels of caffeine to baits could improve ant’s ability to learn its location, improving bait efficacy.

• sample_videos.zip: A subset of the videos used for data extraction. The complete collection of videos is not publicly accessible primarily due to their considerable size (105.35GB). Requests for access to the entire video set are encouraged.
• Preregistration.pdf: The preregistration created for data collection and analysis with justifications for deviations from it.
• OpLan_D1_metadata.csv: Manually collected metadata pertaining to experimental conditions, subjects, and treatments.
• OpLan_D2_DLC_coordinates.zip: Cartesian coordinates obtained from DeepLabCut for each of the videos analysed.
• OpLan_C1_reproject_coordinates.py: Python code used to standardise the ants' coordinates by ensuring the same corner of the A4 platform was used as the origin of the cartesian referential of all videos. The known dimensions of the A4 were further used to convert coordinates from pixels to millimetres.
• OpLan_C2_remove_impossibilities.py: Python code used to account for DeepLabCut tracking errors, with any ant movement exceeding two millimetres per frame being considered implausible and subsequently removed.
• OpLan_C3_find_changepoints.py: Python code used to automatically derive the times at which an ant reached and left the reward from the tracking data.
• OpLan_C4_inward_outward_data.py: Python code used to calculate relevant measures for the foodward (inward) and nestward (outward) journey such as journey duration, mean instantaneous speed and path tortuosity.
• OpLan_C5_Figure_2.R: R code used to produce the raw elements of Figure 2.
• OpLan_C6_Figure_4.R: R code used to produce the raw elements of Figure 4.
• OpLan_C7_Statistical_Analysis.html: Complete statistical analysis and code for the manuscript.