A novel support vector machine-based 1-day, single-dose prediction model of genotoxic hepatocarcinogenicity in rats

Area of Science:

• Toxicology
• Genomics
• Carcinogenesis

Background:

• Accurate cancer risk assessment requires rapid identification of genotoxic and carcinogenic chemicals.
• Genotoxic hepatocarcinogens (GHCs) pose significant health risks, necessitating reliable detection methods.

Purpose of the Study:

• To develop a rapid, 1-day, single-dose rat model for identifying GHCs.
• To establish a predictive classifier for GHCs using gene expression data.

Main Methods:

• Utilized microarray gene expression data from rats exposed to 58 compounds (5 GHCs) from the Open TG-GATEs database.
• Identified 10 common gene markers responsive to GHCs and built a support vector machine classifier.
• Validated the model using multi-institutional 1-day oral administration studies with quantitative PCR (qPCR) on 64 compounds (23 GHCs).

Main Results:

• The GHC predictive model demonstrated high accuracy in silico.
• qPCR proved an effective alternative to microarray analysis for gene expression profiling.
• The model achieved 91% sensitivity and 93% specificity in multi-institutional validation studies.

Conclusions:

• The 1-day single oral administration model is a reliable and highly sensitive tool for GHC identification.
• This model is anticipated to be valuable for screening and identifying potential GHCs in chemical safety assessments.