Completed the collaborative support of investigators' research: 1) Identifying early indicators of toxicity. Human subjects in a clinical study were administered repeated dosing of 1 gm of acetaminophen (APAP) for seven days and then allowed to recover. Blood was drawn daily and analyzed for gene expression and clinical chemistry. Alanine transaminase (ALT) rises approximately 3.5 days after APAP dosing and peaks around day nine. Subjects were classified according to their response to APAP or administration of a placebo. We analyzed the gene expression data with a discontinuous piecewise mixed linear regression model and tested if the change in gene expression at the break point is equal to zero. We identified genes with expression significantly altered within 24h. In addition, the genes clustered patients who overdosed on APAP apart from controls and predicted the exposure classifications with 100% accuracy. 2) Chemotherapeutics in combination elicit unique transcriptional responses. Combinations of chemotherapies can cause organ toxicity. Bone marrow gene expression data from rats exposed to various combinations of oxaliplatin (Oxali) and topotecan (Topo) at 1, 6 and 24 hours was analyzed to extract patterns and to identify co-expressed genes. The order of drug administration was reversed among dose groups at each time point. Cluster analysis of gene expression profiles at the 1 hr time point revealed distinct patterns of co-expressed genes reflective of the toxicity potentiated by the combination and order of the drugs administered. These results demonstrate the potential for early mRNA patterns derived from target organs of toxicity to inform toxicological risk and molecular mechanisms for agents given in combination. 3) Detection of biological processes that differ by chemical mode of action. Gene expression data acquired from the livers of mice exposed to one of 15 chemicals that fall into one of five modes of action (MOA) was assayed using microarray and RNA-Seq platforms, preprocessed and differentially expressed genes detected that were in common between platforms. Using subtrees from enriched gene ontology biological processes, we labeled clusters according to the term that has the most paths to the root. Each MOA is represented by a distinct biological category. 4) Elucidation of co-expressed genes. Most current tools for RNA-seq data are limited in that they do not reveal correlated genes across multiple groups. We developed EPIG-Seq for extracting patterns and identifying co-expressed genes from RNA-Seq data. Using a correlation metric for count data, dispersion estimation and magnitude of change, EPIG-Seq extracts patterns of co-expressed genes across multiple experimental conditions even when sample sizes are small and in data with inflated zeros.
Showing the most recent 10 out of 49 publications
