抄録
Ecotoxicogenomics has emerged as an attractive approach for clarifying critical molecular events altered by ecotoxicants. The aim of the present study was to develop a prediction method for a narcotic mode of toxic action (i.e., baseline toxicity) of chemicals by analyzing the differentially expressed genes in medaka (Oryzias latipes) following exposure. Male medaka were exposed to 15 chemicals including narcotic and reactive compounds for one hour. After exposure, the gene expression profiles of the whole bodies were analyzed using a custom microarray with 36,398 nonredundant transcripts. A total of 576 genes differentially expressed between medaka exposed to narcotic and reactive chemicals were statistically selected. Then, a gene network analysis was performed for their human orthologs in order to find gene sets discriminating between these two toxicant classes. The highest ranked network of genes contained 22 genes, which are involved in calpain-mediated signaling. Subsequently, support vector machine was employed to develop prediction model of baseline toxicity. The overall prediction accuracy by means of leave-one-out crossvalidation was 86.7%. These results indicate that one-hour exposure assays for ecotoxicant classification using transcript profiling in medaka have great promise.
