We show that this approach enables the development of gene expression predictors from genes directly related to biological processes that a conventional single-gene level predictor does not identify. We apply this approach to pinpoint the biological hallmarks of response of two different vaccines, and shows that signatures consistent with proliferating B cells predict antibody response to influenza vaccination. We began by analyzing PBMC microarray data from individuals vaccinated with the yellow fever virus vaccine (YF-17D). YF-17D is a highly potent vaccine that induces a robust interferon gene response in postvaccination PBMC samples [4-6]. In this small data set, our goal was not to identify predictors

of response, but rather to test whether a gene set based analytical approach could recover known biological features of the effect of YF-17D vaccination such as the interferon response. To identify sets of genes PLX-4720 manufacturer — rather than individual genes — that were elicited by YF-17D, we used a variant of gene set enrichment analysis (GSEA) [13]. GSEA is an analytic approach that tests for enrichment of a priori set of genes in a second, rank-ordered list of genes. Such a rank-ordered list of buy Lumacaftor genes is usually created by comparing

the average expression values of genes in a group of microarray samples to those in a control group. Enrichment is measured by the degree of overrepresentation of the set of genes of interest at the top (or bottom) of the rank-ordered list. Because we wanted to test for enrichment of gene sets in individual samples from vaccinated patients (rather than in a group of samples from vaccinated subjects), we used a single sample version of GSEA (ssGSEA) [14]. In this approach, gene sets are tested for enrichment in the list of genes in a single sample ranked by absolute expression rather than by comparison with another sample. We analyzed Affymetrix expression profiles of 15 individuals obtained prevaccination (day Vitamin B12 0) and 7 days following vaccination (day 7). We used ssGSEA to test each sample for enrichment of signatures in a compendium ∼3000

gene sets that have been collected by curation of published microarray studies, or are present in pathway databases such as Reactome (described in the Materials and methods) [11]. We found that ∼900 gene sets were significantly (FDR < 0.25) enriched in the day 7 postvaccine samples (Fig. 1A), suggesting marked differences in gene expression profile following vaccination with YF-17D. To identify whether the gene sets represented similar biological processes, we tested the gene sets for similarity to each other using two approaches. First, we used the DAVID annotation tool [15] to categorize the genes in each gene set and found that the majority of gene sets were strongly associated with the interferon or inflammatory response (Fig. 1A and Supporting Information Table 1).