Table 2.
Classification of individual gene expression profiles by a support vector machine (SVM)-based predictive model.
| Compound | Dose | Classification (n animals) | Classification status (% animals) | |
|---|---|---|---|---|
| Group 1 | Control | – | Control (5) | Unique (60%) |
| Fits into no class (40%) | ||||
| Group 2 | Compound A | 300 mg/kg | Steatotic (5) | Unique (60%) |
| Ambiguous (40%) | ||||
| Group 3 | Compound A | 1500 mg/kg | Steatotic (5) | Unique (100%) |
| Group 4 | Compound B | 300 mg/kg | Control (2) | Fits into no class (100%) |
| Cholestatic (1) | Fits into no class (100%) | |||
| Steatotic (2) | Fits into no class (100%) | |||
| Group 5 | Compound B | 1000 mg/kg | Control (4) | Unique (75%) |
| Fits into no class (25%) | ||||
| Cholestatic (1) | Fits into no class (100%) |
Gene expression profiles from the livers of control and treated rats were classified using the SVM Hepatotox_SVM_OVA_Version l. These were compared with predictive SVM-based models created from our database of liver gene expression profiles for the following hepatotoxic categories: control/nontoxic, direct acting, peroxisome proliferation, cholestasis, and steatosis.