Discovering Chemical Activity Networks-Predicting Bioactivity Based on Structure

点击翻译按钮获取中文摘要

PROJECT SUMMARY NIEHS has established Predictive Toxicology as a strategic goal for advancing environmental health sciences. The overarching goal of this RIVER proposal is to predict animal toxicity of chemicals based on their structure. My team and I will expose millions of zebrafish embryos to a library of 10,000 synthetic chemicals across wide concentration ranges. If a chemical shows signs of bioactivity, we will systematically analyze whole animal gene expression changes before the phenotype appears. We will formulate hypotheses about which biomolecular targets the chemicals attacked initially and which pathways led to the observed endpoint. To test those hypotheses, we will edit the zebrafish genome via CRISPR/Cas9 to knock out or over-express critical genes, to discover the ones causally related to the chemical phenotypes. These studies will be highly relevant to human health. Zebrafish possess fully integrated vertebrate organ systems that perform the same functions as their human counterparts and demonstrate well-conserved physiology. Eighty-four percent of the genes that participate in human disease also exist in zebrafish. Zebrafish studies provide a fast, inexpensive way to screen a large volume of chemicals, generate rich hypotheses for drug development, and prioritize candidates for toxicity studies with mammals and human cell cultures. We will compare our results with those of human cell culture studies to clarify the strengths and weaknesses of each method and to reduce the uncertainty associated with applying zebrafish results to human biology. We will post our experimental results in a public database that explains which of the 10,000 Tox21 chemicals are bioactive, which initial targets they strike, and which pathways lead to which endpoints in embryonic and juvenile zebrafish. This information will enable green chemists to detoxify products by substituting a biologically inactive molecule. It will help toxicologists and risk assessors to prioritize chemicals for expensive experiments with rodents and human cell cultures. It will give pharmaceutical scientists thousands of new data points upon which to develop hypotheses about how to modulate a given gene target or activate a given pathway. We will use machine-learning-based chemoinformatic approaches to analyze our zebrafish data and infer the relationship between the structure of a chemical and its biological activity. Our rich data about chemical activity networks will advance the scientific community’s understanding of linkages between chemical exposure and phenotypes. Our work will enable scientists to predict whether a chemical will be biologically active, what target it will act upon, and what networks it will perturb, solely on the basis of its structure. It will enable scientists to reduce, refine, and replace experiments with animals, including zebrafish, and to predict chemical activity networks with computers.