Large-scale validation of 46 invasive species assays using an enhanced in silico framework

The need for widespread occurrence data to inform species conservation has prompted interest in large, national-scale environmental DNA (eDNA) monitoring strategies. However, targeted eDNA assays are seldom validated for use across broad geographic areas. Here, we validated 46 new and previously published probe-based qPCR assays targeting invasive species throughout the continental United States. We drew upon current taxonomies, range maps, publicly available sequences, and tissue archives to evaluate all potentially sympatric confamilial species and genetically similar extrafamilial taxa. Out of 5276 unique assay-nontarget taxon combinations, we were able to test 4206 (80%). We characterized levels of validation and specificity for each of eight federal geographic regions and provided an online tool with state-level information, as well as detailed assay descriptions in an appendix. Specificity testing benefited from extensive use of eDNAssay-a machine learning classifier trained to predict qPCR cross-amplification-which we found to be 96% accurate in 649 unique tests that underwent paired in silico and in vitro testing. Predictions of assay specificity (the true negative rate) were 98-100% accurate, depending on the classification threshold used. This work provides both an immediate resource for invasive species surveillance and demonstrates an enhanced in silico, geographically subdivided validation framework to aid in future large-scale validation efforts. It can be challenging to ensure that eDNA assays are specific to their intended targets, particularly when biodiversity is high and the geographic area is large. Here, we use an accurate machine-learning-based PCR model to streamline validation of 46 qPCR assays targeting invasive species throughout the continental United States. We present an overview of our validation framework and trends in assay performance, along with detailed records of validation for each assay in an appendix.image