Skip to main content
. 2018 Apr 28;92(6):2055–2075. doi: 10.1007/s00204-018-2213-0

Fig. 5.

Fig. 5

Our imaging-based assay is more predictive than cell count or viability assays. a Percentages of viable cells with respect to the solvent controls for all the 33 reference chemicals (red = pulmonotoxic, blue = non-pulmonotoxic) and 17 test chemicals (green). The values were measured using a standard cell-viability assay with BEAS-2B cells exposed to the chemicals for 72 h. Our final assay based on F1 and F2 and trained on all the reference chemicals was applied to all the reference and test chemicals (solid lines = predicted to be positive, dashed lines = predicted to be negative). b Test balanced accuracy, sensitivity, and specificity values of the three indicated assays in classifying the reference chemicals estimated using a tenfold cross-validation procedure. c Schematic showing the four major pathways activated by our reference and test chemicals. (Color figure online)