Understanding the mediators of innate immunity requires interrogating compendia of knowledge generated from high-throughput technologies 
. In this study we analyzed macrophage activation across a broad-spectrum of innate immunity stimuli; inert nanoparticle exposure, TLR agonists, and Salmonella
infection. Using a combination of bioinformatics techniques we determined a highly focused group of candidate genes for further experimental investigation. The topology of the inferred macrophage regulatory network was used to identify many of these important genes. We showed that the bottlenecks of the network are significantly enriched in known targets of pathogens. This finding can be compared to those reported in the human protein-protein interaction network 
and our previous findings that bottlenecks were enriched in virulence essential genes 
We next examined genes that are differentially regulated in response to multiple stimuli and identified a subset of genes that are differentially expressed under all conditions examined and occupy a highly central location in the inferred network. Interestingly these genes are highly enriched in conserved homologs and pathogen targets, indicating that they are biologically significant in macrophage activation. This group, the macrophage core response module, which encompasses 38 genes including chemotactic cytokines (Ccl3, Ccl4, Cxcl2), transcription factors (Fos/Jun AP-1 complex, Egr1 and 2, and Mafb), apoptosis (Ddit3 and Gadd45b), steroid biosynthesis (Fdft1, Sc4mol, and Scd1), other immune response-related genes (Ifit1, Ifit2, Mx1, Mx2, Oas2, Oasl1 and Ptsg1 [Cox1]) and a number of genes with unknown roles in immune response (see ). Previously, Ramsey, et al. (2008) 
analyzed a macrophage compendium (also used as part of the present study) using a variety of approaches and described two ‘core early response’ clusters that overlap with our module significantly (~50% shared genes). Furthermore, through motif enrichment they found that genes in these clusters were enriched in AP1, JUN, CREB, ATF, EGR, and PPARA binding sites, indicating that components of our core response module may be regulated by the Fos/Jun AP1 complex, and the Egr1 and 2 transcription factors that are in the module as well.
To represent the dynamics of the genes in this module and predict the regulatory influences governing their expression we developed a predictive model. By describing the regulatory network of the core response module in a machine learning algorithm 
, we were able to predict gene expression on a new data set. Multivariate regression techniques have been applied to model data in prokaryotes 
or yeast 
, and here we have successfully applied this method to model data from a mammalian system. Our resulting model accurately predicts the behavior of the core response module in combinations of treatments and genetic backgrounds in a cross-validation approach. Furthermore, the model can accurately predict the expression of a subset of these genes in macrophages responding to nanoparticle exposure, which induces a very different response than the TLR pathway.
Our predictive model identifies a number of regulatory influences that provide the basis for further experimental investigation. Core response module subcluster 1, which is enriched in transcription factors like Egr1/2 and Fos/Jun, is predicted to be regulated by Nfkß2, a component of the alternative Nfkß pathway 
, and negatively regulated by Purb and Zkscan1, intriguingly neither of which has a demonstrated role in innate immunity. Using the Metacore program (GeneGO, St. Joseph, MI) that has a curated database of known regulatory relationships, we found that six of the 14 members of subcluster 1 were known to be regulated by one or more of the regulators inferred in our analysis. Subcluster 3 is composed of many interferon regulated genes, and is predicted to be regulated by Irf4 and Nr2f6. Irf4 may be involved in alternative macrophage activation by IL-4 
, and is known to regulate Ifit2 
, but the function of Nr2f6 in macrophages is unknown. Interestingly, ISGF3, a regulatory complex composed of Stat1, Stat2, and Irf9, is known to be a primary regulator of the interferon response, but is not identified by our analysis. This is likely due to the fact that the activity of this complex is not closely tied to the expression levels of its component genes, requiring phosphorylation and assembly of the protein complex itself. This limitation does not refute the predictions made by our approach since it has been shown that regulation of the interferon response is complicated and involves multiple redundant pathways 
. Finally, cluster 4 is composed of three genes, Ptgs1 (Cox-1), and the cytokines Ccl4 and Cxcl2. These genes are highly upregulated under nearly all stimuli examined and are predicted in our model to be regulated by Nfkß and Rela, the complex responsible for primary activation of the inflammatory response. Strikingly, all three of these genes are known to be regulated by Nfkß, supporting our inferred model. Of the other predicted regulatory influences Nfix has no known immune response functions, but St18 is a known regulator of the proapoptotic response 
, which is related to inflammation.
Our identification and characterization of the core response module suggests that it plays an important role in macrophage activation. The known functions of some of its members, for example AP1, the module's central location in the inferred network, and its preferential targeting by pathogens suggest that it may be an early mediator of downstream functions, possibly as a checkpoint of progression to apoptosis or inflammation. Our analysis suggests that lipid and cholesterol biosynthesis pathways are an important response, a portion of which is triggered by a general response to particles and possibly not through classical TLR pathways, though further investigation is needed to confirm this observation. A future direction is to investigate the downstream functions that the module may be involved in regulating and determine how pathogen proteins may alter this regulation to promote virulence.
Bioinformatic studies of macrophage response to TLR agonists and to bacterial infection using a compendia of transcriptomic data have been published previously 
, and have reported similar core response sets of genes that are much larger than ours. Our study is the first to compare these responses with those elicited by inert manufactured nanoparticles; deducing a more concentrated subset of regulators. We identified lipid and cholesterol biosynthesis pathways as being potentially responsive to particles including nanoparticles and live bacteria. The core response module appears to be highly relevant to macrophage activation as we showed by training on TLR agonist and Salmonella
infection and very accurately predicting the dynamic behavior of gene expression under nanoparticle exposure. This analysis is significant because it shows that although much of the macrophage response differs for nanoparticles, a set of genes is regulated by all three kinds of responses, and this set, our core response module, seems to be a very important component of macrophage activation.