Email updates

Keep up to date with the latest news and content from BMC Genomics and BioMed Central.

This article is part of the supplement: The 2010 International Conference on Bioinformatics and Computational Biology (BIOCOMP 2010): Genomics

Open Access Research article

Predicting adverse side effects of drugs

Liang-Chin Huang1, Xiaogang Wu12 and Jake Y Chen12*

Author affiliations

1 School of Informatics, Indiana University, Indianapolis, IN 46202, USA

2 MedeoLinx, LLC, Indianapolis, IN 46280, USA

For all author emails, please log on.

Citation and License

BMC Genomics 2011, 12(Suppl 5):S11  doi:10.1186/1471-2164-12-S5-S11

Published: 23 December 2011

Abstract

Background

Studies of toxicity and unintended side effects can lead to improved drug safety and efficacy. One promising form of study comes from molecular systems biology in the form of "systems pharmacology". Systems pharmacology combines data from clinical observation and molecular biology. This approach is new, however, and there are few examples of how it can practically predict adverse reactions (ADRs) from an experimental drug with acceptable accuracy.

Results

We have developed a new and practical computational framework to accurately predict ADRs of trial drugs. We combine clinical observation data with drug target data, protein-protein interaction (PPI) networks, and gene ontology (GO) annotations. We use cardiotoxicity, one of the major causes for drug withdrawals, as a case study to demonstrate the power of the framework. Our results show that an in silico model built on this framework can achieve a satisfactory cardiotoxicity ADR prediction performance (median AUC = 0.771, Accuracy = 0.675, Sensitivity = 0.632, and Specificity = 0.789). Our results also demonstrate the significance of incorporating prior knowledge, including gene networks and gene annotations, to improve future ADR assessments.

Conclusions

Biomolecular network and gene annotation information can significantly improve the predictive accuracy of ADR of drugs under development. The use of PPI networks can increase prediction specificity and the use of GO annotations can increase prediction sensitivity. Using cardiotoxicity as an example, we are able to further identify cardiotoxicity-related proteins among drug target expanding PPI networks. The systems pharmacology approach that we developed in this study can be generally applicable to all future developmental drug ADR assessments and predictions.