You are here:
Predicting Molecular Initiating Events from High Throughput Transcriptomic Screening using Machine Learning (ISMB 2021)
Citation:
Bundy, J., R. Judson, I. Shah, A. Williams, Chris Grulke, AND L. Everett. Predicting Molecular Initiating Events from High Throughput Transcriptomic Screening using Machine Learning (ISMB 2021). Intelligent Systems for Molecular Biology (ISMB), Virtual, NC, July 25 - 30, 2021.
Impact/Purpose:
This is an abstract to be submitted to the 29th conference on Intelligent Systems for Molecular Biology (ISMB), which is an international meeting covering bioinformatics and computational biology (July 2021, virtual). The abstract and subsequent poster or presentation will communicate our work on developing machine learning based methods for predicting molecular initiating events from gene expression data sets.
Description:
The advent of high-throughput transcriptomic screening technologies has resulted in a wealth of gene expression signatures associated with chemical treatment. One resource is the Library of Integrated Network-Based Cellular Signatures (LINCS) dataset, spanning ~20k chemical perturbagens. Such data sets offer utility for the prediction of molecular initiating events (MIEs) induced by chemical exposure. The development of these predictive methods is relevant to U.S. EPA’s focus on increasing efficiency in chemical screening using new approach methodologies (NAMs). To ascertain the utility of high-throughput transcriptomic data for hazard identification, we trained binary classifiers to predict MIEs from gene expression. Classifiers were trained using three training feature types, six classification algorithms, and two cell-types (MCF7 and PC3). This analysis identified high performance classifiers for 9 MIEs in MCF7 cells based on predictions excluded from classifier training. MIEs modeled with dissimilar accuracies between cell lines, such as estrogen receptor activation, were found to correspond to targets that have different baseline expression in MCF7 and PC3 cells. Linking MIE labels with gene expression compendia can produce models with predictive value and can also help identify suitable cell lines for screening particular MIEs. This abstract does not necessarily reflect US EPA policy.