The mouse embryonic stem cell test (EST) is a 10-day screen for teratogenic potential developed to reduce animal use for embryotoxicity testing of chemicals (Spielmann, 2005; Spielmann et al., 1997). In this study, we used the cytotoxicity IC(50) values and transcriptional expression changes as primary endpoints in a shorter 4-day version of the EST, the molecular embryonic stem cell assay. Mouse D3 embryonic stem cells were used for cytotoxicity assessment (monolayers) or grown as embryoid bodies in low attachment plates for transcriptional profiling. Sixty-five compounds with known in vivo teratogenicity (33 teratogens and 32 nonteratogens) were evaluated to develop a model for classifying compounds with teratogenic potential. The expression of 12 developmentally regulated gene targets (nanog, fgf5, gsc, cd34, axin2, apln, chst7, lhx1, fgf8, sox17, foxa2, and cxcr4) was measured following exposure of embryoid bodies to a single compound concentration (0.1 × the cytotoxicity IC(20)) for 4 days. In the decision-tree model, compounds with IC(50) values < 22 µM were categorized as teratogens, whereas compounds in the two groups with IC(50) values between 22-200 µM and > 200 µM were categorized as teratogens if ≥ 8 and 12 genes, respectively, were deregulated by at least 10%. Forty-seven of 65 compounds of the training set were correctly identified (72% total concordance). In a test set of 12 additional compounds (5 teratogens, 7 nonteratogens), 10 were correctly classified by this approach (83% concordance). The false positive rate in the training and test sets was 24 and 0%, respectively, indicating that this assay has potential to identify teratogens.