Cross-site comparison of gene expression data reveals high similarity - Genomics and Risk Assessment: Mini-Monograph

Environmental Health Perspectives, March 15, 2004 by Tzu-Ming Chu, Shibing Deng, Russ Wolfinger, Richard S. Paules, Hisham K. Hamadeh

Consistency and coherence of gene expression data across multiple sites depends on several factors such as platform (oligo, cDNA, etc.), environmental conditions at each laboratory, and data quality. The Hepatotoxicity Working Group of the International Life Sciences Institute Health and Environmental Sciences Institute consortium on the application of genomics to mechanism-based risk assessment is investigating these factors by comparing high-density gene expression data sets generated on two sets of RNA from methapyrilene (MP) experiments conducted at Abbott Laboratories and Boehringer-Ingdheim Pharmaceuticals, Inc. using a single platform (Affymetrix Rat Genome U34A GeneChip) at seven different sites. This article focuses on the evaluation of data quality and statistical models that facilitate the comparison of such data sets at the probe level. We present methods for exploring and quantitatively assessing differences in the data, with the principal goal being the generation of lists of site-insensitive genes responsive to low and high doses of MP. A combination of numerical and graphical techniques reveals important patterns and partitions of variability in the data, including the magnitude of the site effects. Although the site effects are significantly large in the analysis results, they appear to be primarily additive and therefore can be adjusted in the statistical calculations in a way that does not bias conclusions regarding treatment differences. Key words: cross-site comparison, gene expression, hepatotoxicity, ILSI, toxicogenomics. Environ Health Perspect 112:449-455 (2004). doi:10.1289/txg.6787 available via http://dx.doi.org/[Online 15 January 2004]

**********

Advancing the general knowledge base of mechanisms and markers of hepatotoxicity is of great interest to all parties involved in this consortium. The Hepatotoxicity Working Group of the International Life Sciences Institute (ILSI) Health and Environmental Sciences Institute (HESI) Committee on the Application of Genomics to Mechanism-Based Risk Assessment is investigating these factors by comparing high-density gene expression data sets generated on two sets of RNA from two independent in vivo experiments where rats were dosed with methapyrilene (MP) conducted at either Abbott Laboratories (site A; Abbott Park, IL) or Boehringer-Ingelheim Pharmaceuticals, Inc. (BIPI; site B; Ridgefield, CT).

Most microarray studies are designed with large "p" (number of genes) and small "n" (number of arrays) characteristics. Two issues of concern arise when investigators work with data having such characteristics. The first issue is of statistical inference power where the aim is to minimize both false-positive and false-negative rates. Increasing sample size (i.e., n) can afford better statistical inference power; however, this remedy is often cost prohibitive. The second issue arises in the attempt to address the first concern by increasing sample size by incorporating data sets generated at disparate sites and times. Thus, the second concern is about the consistency of such data sets generated across multiple sites and whether the same or similar conclusions can be drawn. An across-site microarray study can be useful for addressing this issue, which is another way to increase sample size. Conceptually, the complexities among data generated across different sites are higher than those of data generated within one site. The above two issues are related; however, the second one may be more general and of increasing concern as microarray data sets become increasingly available and the desire to compare and contrast across studies increases.

Male Sprague-Dawley rats [CRL: CD(SD)IGS VAF/Plus] (Charles River Laboratories, Kingston, NY) approximately 6-7 weeks of age were assigned to nine study groups (four rats/group) and dosed by garage for 1, 3, or 7 days with water (vehicle), 10 mg/kg/day MP, or 100 mg/kg/day MP (Figure 1). Dose selection was based on published and unpublished studies; the high dose of MP was chosen to yield hepatotoxicity, and a nontoxic low dose was selected. In general, the BIPI study yielded more hepatotoxicity than the study conducted at Abbott Laboratories, as defined by clinical pathology parameters and microscopic examinations of hematoxylin and eosin-stained liver sections. No significant histopathological alterations were observed in livers of rats treated with 10 mg/kg/day MP at the 1- and 3-day time points compared with alterations in livers of the control groups. In comparison, MP treatment with 10 mg/kg/day for 7 days resulted in minimal portal mononuclear infiltrates, minimal hepatocellular periportal necrosis, and minimal microvesicular hepatocellular vacuolization. At the 100-mg/kg/day dose, all rats showed early minimal mononuclear portal infiltrates, minimal hepatocellular periportal necrosis, and mild to moderate periportal microvesicular vacuolization at 1 and 3 days of exposure. The severity of the lesions increased at day 7, and mild hyperplasia became evident. In addition, in the 100-mg/kg/day MP dose group at 7 days of exposure, moderate mononuclear portal infiltrates were noted, and the number of enlarged periportal hepatocytes with microvesicular vacuolization increased. The severity of hepatocellular periportal necrosis at the 7-day time point also increased, accompanied by increased numbers of hepatocellular mitotic figures. Bile duct hyperplasia was observed in animals in the 100-mg/kg/day dose group at 3 and 7 days. Minimal bile duct hyperplasia was seen at the 3-day time point and increased in severity to mild by 7 days. Levels of alanine aminotransferase, aspartate aminotransferase, and sorbitol dehydrogenase increased in high-dose animals in a time-dependent manner. Total bilirubin tended to be elevated in the high-dose group with continued dosing. All the above parameters were reflective of liver toxicity.