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).