We describe the development of computational models that predict activity in a repeat-dose zebrafish embryo developmental toxicity assay using a combination of physico-chemical parameters and in vitro (human) assay measurements. The data set covered 986 chemicals including pesticidal active ingredients, pharmaceuticals, industrial chemicals, food additives and consumer products. Approximately 50% of chemicals caused developmental defects, up to and including embryo death. The most parsimonious, but highly predictive, machine learning model used a single physicochemical parameter (logP) and a set of the in vitro assays that measure cell stress and cytotoxicity. The resulting model predicted developmental toxicity in zebrafish with ~80% balanced accuracy. Probability of developmental toxicity peaks at logP~4.3 and drops off at higher and lower values. The stress model includes in vitro assays for cytotoxicity, oxidative stress, apoptosis, and microtubule and mitochondrial disruption. The inclusion of specific gene targets and processes did not improve that accuracy of the classification model (active / inactive). Further analyses, however, examining relative potency of the active compounds (those causing developmental defects) identified a number of molecular targets or pathways that help predict which chemicals can cause developmental toxicity in zebrafish embryos at concentrations below where stress pathways are activated, sometimes referred to as “excess toxicity”. These include specific activity against mitochondria, microtubules, endocrine pathways (estrogen, androgen, glucocorticoid, progesterone and thyroid), and ion channels. Taken together, these data describe a predictive toxicity model for the developing zebrafish embryo. More broadly, they highlight the utility in applying machine learning approaches to identify discrete suites of chemical features and high throughput in vitro assays that collectively predict developmental toxicity in vivo. The findings from this study will be applied to understanding potential risks to human health, but will also help identify chemical activities that are likely to be toxic to developing fish embryos and other ecological species. This abstract does not necessarily reflect U.S. EPA policy.