New investigative tools such as gene expression profiling have begun to be applied to the problem of predicting vaccine response [2]. Most of these approaches have assayed vaccine-induced changes in gene expression in the PBMC compartment, a bellwether of changes at distant vaccine sites. Two studies have shown that changes in the expression of small numbers of genes in PBMC gene expression profiles a few days after vaccination predict the subsequent magnitude of the immune response measured several weeks later [3, 4]. These studies suggest that gene expression profiles from PBMC samples in vaccinated subjects can selleck inhibitor provide predictors of the

vaccine response. Such approaches would be especially useful both as tools to identify new biological features associated with vaccine response, and as correlates of immunity for the development of new vaccines. However there are two significant challenges to developing gene expression based predictors of clinical outcome following vaccination. First, the extent of biological change in PBMCs caused by direct interaction with the vaccine and PBMCs would be expected to be small. Although live attenuated vaccines such as those developed against yellow fever (YF-17D) are known to replicate

systemically and induce readily detectable interferon responses [4-6], nonreplicating subunit vaccines such as those against influenza would be expected to have a much smaller effect GDC-0068 on the transcriptional profile of PBMCs. Thus the selection of individual genes that are strongly associated with response to vaccination can be difficult. The second challenge is that the biological meaning of gene expression based predictors is often hard to determine [3, 4]. One reason for this is that the analytical approaches to identify predictive genes are often different from those used to discover biological mechanisms evident in gene expression data. Predictive genes are selected on statistical rather than biological grounds [7], which tends to divorce the identity of the predictive genes from an understanding

of their role in vaccine Abiraterone price biology [8]. To address these limitations, we applied an approach to developing predictors of vaccine outcome from PBMC gene expression profiles following vaccination that has been used in other domains, e.g. stratifying cancer patients, but is novel to immunology. Rather than building a predictive model based on single differentially expressed genes, we used sets of coordinately regulated, biologically informative gene sets as predictive features in individual samples [9, 10]. As a source of gene sets, we use a compendium of signatures extracted from the published literature and from expert curation [11]. These signatures represent phenotypes of defined cell states and biological perturbations, providing specific biological contexts with which to interpret the predictive models.