Evaluating external datasets using CoHEC models.

MOA prediction accuracy and performance when applying our methods to the data from Zoffmann et al. 2019 and Zampieri et al. 2018 and the methods from Hutter et al. 2004 on our dataset. In all cases, LCOCV was used for evaluating model performance for each individual observation (e.g. pairwise DGE profile), each cross-validation set (e.g. held out teixobactin), and using various majority voting schemes (see Materials and Methods). CPD is an abbreviation for compound. *Indicates protein-synthesis sub-MOA classification (30S/50S).