We then tested each subject’s vaccine response for enrichment of gene sets from the same database collection as used before using ssGSEA and identified the gene sets most differentially enriched in
the high responders compared with the low responders. We found 13 gene sets significantly associated with a high HAI response to vaccine (FDR < 0.25) (Fig. 2A). The number of gene sets and degree of enrichment of gene sets correlated with TIV antibody response was lower than what we observed in the comparison of pre- and post-YF-17D vaccination. This suggests that the biological “signal” associated with influenza vaccine response is less pronounced than the effect of buy R428 vaccination with YF-17D. The gene sets that were enriched in responders were from a wide array of studies and sources (Supporting Information Table 2) and the genes in most gene sets were nonredundant (Supporting Information Fig. 2), suggesting that the gene sets represented diverse biological processes. However, using a constellation Fulvestrant concentration plot, we found two distinct but connected clusters of gene sets (Fig. 2B). We used DAVID annotation as a tool to provide secondary annotation for the two clusters
of genes and found that one cluster (indicated by the orange arc) was strongly enriched for immunoglobulin and complement genes. The second cluster (indicated by the purple arc) was strongly enriched for genes associated with proliferation (Supporting Information Table 3). Only a subset of proliferation-related gene sets contained in MSigDB enriched in responders (Supporting Information Fig. 3) suggesting that the proliferation signature present in vaccine responders is not shared by all tissue types. Alternatively, other proliferation-related gene sets in the compendium may also entrain other biological responses not present in vaccine responder expression profiles. We reasoned that if these clusters of highly connected gene sets enriched in samples from vaccine responders represented bona fide biological processes, then
the genes shared by each of these clusters should be overrepresented for physically interacting genes. To test this, we projected the genes Anacetrapib found in the gene set clusters into InWeb [18] a curated protein–protein interaction network (PPI; Fig. 2C and D). We found that there was a high degree of physical connectivity between the component genes of the antibody gene cluster (p = 10−3), and between the genes in the proliferation cluster (p = 10−2) (Fig. 2C and D). This suggests that the clusters of enriched gene sets found in responders represented coordinated upregulation of genes in functional networks. We confirmed these findings using a second, independent source of gene sets, described by Chaussabel et al.