Classification of Developmental Toxicants in a Human iPSC Transcriptomics-Based Test

Despite the progress
made in developmental toxicology, there is
a great need for in vitro tests that identify developmental toxicants
in relation to human oral doses and blood concentrations. In the present
study, we established the hiPSC-based UKK2 in vitro test and analyzed
genome-wide expression profiles of 23 known teratogens and 16 non-teratogens.
Compounds were analyzed at the maximal plasma concentration (Cmax) and at 20-fold Cmax for a 24 h incubation period in three independent experiments. Based
on the 1000 probe sets with the highest variance and including information
on cytotoxicity, penalized logistic regression with leave-one-out
cross-validation was used to classify the compounds as test-positive
or test-negative, reaching an area under the curve (AUC), accuracy,
sensitivity, and specificity of 0.96, 0.92, 0.96, and 0.88, respectively.
Omitting the cytotoxicity information reduced the test performance
to an AUC of 0.94, an accuracy of 0.79, and a sensitivity of 0.74.
A second method, which used the number of significantly deregulated
probe sets to classify the compounds, resulted in a specificity of
1; however, the AUC (0.90), accuracy (0.90), and sensitivity (0.83)
were inferior compared to those of the logistic regression-based procedure.
Finally, no increased performance was achieved when the high test
concentrations (20-fold Cmax) were used,
in comparison to testing within the realistic clinical range (1-fold Cmax). In conclusion, although further optimization
is required, for example, by including additional readouts and cell
systems that model different developmental processes, the UKK2-test
in its present form can support the early discovery-phase detection
of human developmental toxicants.