You are here:
Predicting Drug-Induced Liver Injury using Transcriptomics Data
Citation:
Shah, I. Predicting Drug-Induced Liver Injury using Transcriptomics Data. Presented at Intelligent Systems for Molecular Biology/Critical Assessment of Massive Data Analysis, Chicago, IL, July 06 - 10, 2018.
Impact/Purpose:
Despite the application of high-throughput screening as well as large-scale molecular profiling (metabolomics, proteomics and transcriptomics) of in vitro and in vivo models, accurately predicting DILI remains an open problem. Here we investigate a multi-strategy data-driven approach to classify drugs into DILI classes (positive or negative) using Affymetrix transcriptomic profiles from the CMap (v2) database.
Description:
Drug-induced liver injury (DILI) is one of the major causes for acute liver failure and for liver transplantation procedures performed in the US. Approximately half of all drug candidates produce some hepatic effects in preclinical studies, and DILI is also a frequent cause for drug failure in clinical trials. Despite the application of high-throughput screening as well as large-scale molecular profiling (metabolomics, proteomics and transcriptomics) of in vitro and in vivo models, accurately predicting DILI remains an open problem. Here we investigate a multi-strategy data-driven approach to classify drugs into DILI classes (positive or negative) using Affymetrix transcriptomic profiles from the CMap (v2) database. Three main strategies will be used to classify chemicals: (a) connectivity-mapping, (b) machine learning, and (c) pathway-based scoring. First, the training data set will be used to estimate cross-validation accuracy of the different prediction techniques. Second, the testing data will be used for independent validation of each of the approaches (after unblinding). Additional results will be submitted by the CAMDA submission deadline and presented at the ISMB 2018/COSI. This abstract does not reflect EPA policy.