False positives in the vouchered reference DNA pool
Although the FR community included specimens widely sourced to maximize taxonomic diversity, some of these samples were almost certainly exposed to unknown fish and other metazoan contaminants throughout the supply chain (e.g., market samples). Therefore, we used the vouchered samples and the VR DNA pool to test data decontamination and quantify false positive detections (i.e., detection of taxa not added to our pool).
For all markers, the analysis returned several false positives for the VR pool libraries. The species occupancy modeling and Bray-Curtis data decontamination procedure removed some of these ASVs that corresponded to species known to be present in the laboratory where DNA was extracted, and libraries prepared (Fig. 3). Certain additional taxa may have accompanied vouchered specimens through logical mechanisms, such as parasitic organisms (e.g., Kudoa sp .). Species occupancy detection modeling (SODM) removed 18 ASVs with low estimated probability of occupancy, 11 of which corresponded to species-level taxonomy. Another five false positives were filtered by removing sample replicates with Bray-Curtis dissimilarity scores >0.49 (Fig. 3; SI Table S2). Remaining false positives ranged from 0-7 per locus (mean = 2.9) and included taxa known to be present in the lab, including species added to the FR DNA pool. When tallying only unknown contaminant sources, the number of false positives was reduced to 0-6 per locus (mean = 2.4; Fig. 3).