Transcriptional networks characterize ventricular dysfunction after myocardial infarction: a proof-of-concept investigation

J Biomed Inform. 2010 Oct;43(5):812-9. doi: 10.1016/j.jbi.2010.05.012. Epub 2010 May 23.

Abstract

There is currently no method powerful enough to identify patients at risk of developing ventricular dysfunction after myocardial infarction (MI). We aimed to identify major mechanisms related to ventricular dysfunction to predict outcome after MI. Based on the combination of domain knowledge, protein-protein interaction networks and gene expression data, a set of potential biomarkers of ventricular dysfunction after MI was identified. Here we propose a new strategy for the prediction of ventricular dysfunction after MI based on "network activity indices" (NAI), which encode gene network-based signatures and distinguishes between prognostic classes. These models outperformed prognostic models based on standard differential expression analysis. NAI-based models reported high classification accuracy, with a maximum area under the receiver operating characteristic curve (AUC) of 0.75. Furthermore, the classification capacity of these models was validated by performing evaluations on an independent patient cohort (maximum AUC=0.75). These results suggest that transcriptional network-based biosignatures can offer both powerful and biologically-meaningful prediction models of ventricular dysfunction after MI. This research reports a new integrative strategy for identifying transcriptional responses that characterize cardiac repair and for predicting clinical outcome after MI. It can be adapted to other clinical domains, such as those constrained by small molecular datasets and limited translational knowledge. Furthermore, it may reflect clinically-meaningful synergistic effects that cannot be identified by standard analyses.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Biomarkers / analysis
  • Cluster Analysis
  • Computational Biology / methods
  • Databases, Genetic
  • Decision Support Systems, Clinical*
  • Gene Expression Profiling
  • Gene Regulatory Networks*
  • Humans
  • Medical Informatics / methods*
  • Myocardial Infarction / genetics*
  • Myocardial Infarction / metabolism
  • Myocardial Infarction / physiopathology
  • Prognosis
  • Protein Interaction Mapping
  • Reproducibility of Results
  • Ventricular Dysfunction / diagnosis
  • Ventricular Dysfunction / genetics*
  • Ventricular Dysfunction / prevention & control

Substances

  • Biomarkers