The U.S. Geological Survey and the U.S. Environmental Protection Agency have assessed contaminants in 38 streams across the U.S., using an extensive suite of target-chemical analysis methods along with a variety of biological effects tools. Here we report zebrafish liver (ZFL) cell-culture based NMR metabolomic analysis of these split stream samples. We used this untargeted approach to evaluate the sites according to overall impact on the ZFL metabolome and found that neither the total number of organics detected at the sites, nor their cumulative concentrations, were good predictors of these impacts. Further, we used partial-least squares regression to compare ZFL endogenous metabolite profiles to values for 455 potential stressors (organics, inorganics, and physical properties) measured in these waters and found that the profiles covaried with at most 280 of the stressors, which were subsequently ranked into quartiles based on the strength of their covariance. While contaminants of emerging concern (CECs) were well represented in the top, most strongly, covarying quartile – suggesting considerable potential for eliciting biological responses at these sites – there was even higher representation of various well-characterized legacy contaminants (e.g., PCBs). These results emphasize the importance of complementing chemical analysis with untargeted bioassays to help focus regulatory efforts on the most significant ecosystem threats.