Curation of prey
To obtain datasets with putative prey only, the assigned reads were subjected to a two-step curation process in R studio (v. 4.1.3). First, the reads were manually filtered based on taxonomy using ‘tidyverse’ functions (Wickham et al., 2019). All Maxillopoda reads were discarded to remove consumer DNA. We acknowledge that maxillopods may compose a food-source for the species studied but the short read-length used to capture prey from partly digested materials, did not allow for distinguishing DNA from maxillopod prey and consumer. Taxa known to interact with copepods (any Copepoda) in symbiosis (parasitism, commensalism and mutualism) were recorded from current literature (Bass et al., 2021; Cleary et al., 2017; Cleary & Durbin, 2016; Zamora-Terol et al., 2021) and used to discard likely non-dietary interactions. By inspection, we discarded several zOTUs assigned to unlikely prey including seed-plants, insects and mammals. Putative contaminants from the marine environment were likewise discarded, notably large gelatinous organisms (Cnidaria, Ctenophora). We acknowledge that also these taxa may have a dietary origin, but we consider it more plausible that most of the sequences originated from the batch sample from which the copepods were picked. Gelatinous organisms are sticky and fragile, and have been suspected of contaminating other studies of copepods using similar methodologies (Cleary et al., 2017). Decontam was used to identify and discard remaining contaminants by comparing the prevalence of putative contaminant zOTUs in real samples and extraction negatives (the ‘prevalence method’, Davis et al., (2018)). Relatively few contaminants (2 and 38 for pilot and full datasets, respectively) were identified and discarded at this stage. The remaining zOTUs were considered putative prey, and were stored with metadata as phyloseq-objects (McMurdie & Holmes, 2013).