Compound Cardiovascular Activity Prediction Using Structural and Genomic Features

点击翻译按钮获取中文摘要

Project Summary Unexpected cardiovascular activity plays a substantial role in therapeutic program failure leading to major loss of patient life and research time. Therefore, it is imperative that steps be taken to understand and predict drug cardiovascular activity. As the age of personalized medicine advances, the use of Gene Expression Signatures (GES) has emerged as a new tool to describe biological processes. These GES consist of the quantitative levels of mRNA expressed in a biological system as a result of a perturbation; however, they lack information about underlying protein structure and function. It has been shown that Structural Gene Expression Signatures (sGES), which integrate protein structure information derived from GES, produce reliable signatures that capture cellular responses to perturbagens such as drugs. Specifically, preliminary results demonstrate how these sGES capture underlying patterns that link compound structure with cardioactivity in cardyomyocites. This preliminary data, combining computational and experimental work, was made possible by the outstanding environment of scientific inquiry nurtured through the long-standing collaboration between this proposal's sponsor and co-sponsor. The goal of this training is to hone skills that bridge the divide between informatics, bench top experiments, and human health as well as develop fluency in scientific thinking that can be applied to a future career as a physician scientist. The mentorship of this proposal's sponsors, the opportunities at ISMMS to learn from diverse collaborations and experiences, as well as the scientific plan outlined in this proposal all contribute to the strength of this training plan to achieve this goal. Specifically, this project will expand the sGES tool through the addition of multiple features derived from structure and function, such as secondary structure and protein disorder, and apply the resulting signatures to cardiovascular activity understanding as well as prediction. The final sGES tool will be made publically available on a web server, which has already been constructed. Next, profiles will be generated for compounds based on their signature, structure, and recorded cardiovascular activity in the FDALabel database. The use of these profiles will be two fold. The first will be as training data for an ensemble learning algorithm, which will predict drug structure from signature and therefore provide a valuable first step toward generating de novo compounds from disease signatures. The second use of these profiles will be to create a map linking chemical structure to cardiovascular activity. The predicted cardiovascular activities from these computational aims will be experimentally validated with cell based cardiotoxicity assays such as the hERG assay, which is a commonly used as a first line screen for cardiovascular toxicity. Ultimately, the completion of this project will result in the development of useful, validated, publically available tools for understanding as well as predicting cardiovascular activity and prepare the investigator to conduct scientific research as a physician scientist.