- Open Access
Coregulation mapping based on individual phenotypic variation in response to virus infection
© Nudelman et al. 2010
- Received: 26 January 2010
- Accepted: 18 March 2010
- Published: 18 March 2010
Gene coregulation across a population is an important aspect of the considerable variability of the human immune response to virus infection. Methodology to investigate it must rely on a number of ingredients ranging from gene clustering to transcription factor enrichment analysis.
We have developed a methodology to investigate the gene to gene correlations for the expression of 34 genes linked to the immune response of Newcastle Disease Virus (NDV) infected conventional dendritic cells (DCs) from 145 human donors. The levels of gene expression showed a large variation across individuals. We generated a map of gene co-expression using pairwise correlation and multidimensional scaling (MDS). The analysis of these data showed that among the 13 genes left after filtering for statistically significant variations, two clusters are formed. We investigated to what extent the observed correlation patterns can be explained by the sharing of transcription factors (TFs) controlling these genes. Our analysis showed that there was a significant positive correlation between MDS distances and TF sharing across all pairs of genes. We applied enrichment analysis to the TFs having binding sites in the promoter regions of those genes. This analysis, after Gene Ontology filtering, indicated the existence of two clusters of genes (CCL5, IFNA1, IFNA2, IFNB1) and (IKBKE, IL6, IRF7, MX1) that were transcriptionally co-regulated. In order to facilitate the use of our methodology by other researchers, we have also developed an interactive coregulation explorer web-based tool called CorEx. It permits the study of MDS and hierarchical clustering of data combined with TF enrichment analysis. We also offer web services that provide programmatic access to MDS, hierarchical clustering and TF enrichment analysis.
MDS mapping based on correlation in conjunction with TF enrichment analysis represents a useful computational method to generate predictions underlying gene coregulation across a population.
- Gene Ontology
- Newcastle Disease Virus
- Enrichment Score
- Jaccard Coefficient
- Putative Transcription Factor Binding Site
Variability in human immune response is a well known fact, confirmed each year when the seasonal influenza arrives. Understanding why individuals show a diversity of immune responses would have important clinical implications. Investigation of the role of functional single nucleotide polymorphisms (SNPs) in immune related genes in the response to challenge by infectious agents is part of the effort to study variability in the human immune response [1, 2]. Here we investigate human variability at a cruder level, the level of gene expression across a group of individuals infected by the same virus. We are looking at correlations and their basis among a number of genes upregulated as a result of the infection. The infection is by Newcastle disease virus (NDV), an avian parainfluenza virus that lacks the tools to disrupt the human innate immune response, and therefore engages it fully. The cells are human dendritic cells (DCs), which are important players in innate immunity and initiators of adaptive immunity through T cell activation. The study of gene correlations in a population, though mathematically similar, is very different from the study of a data set where gene expression levels would be measured at a number of time points after infection for cells from the same individual. Here the variability is found across humans themselves not in the time courses of expression levels [3, 4].
Our data show considerable variability in human immune response for DCs infected by NDV. In order to uncover the underlying patterns of gene expression, we studied gene correlations across the population, with the aim of clustering genes according to some distance measure. There are many distance measures reported in the literature and available via web based computational analysis tools. Our goal was to obtain a combination of distance measure and transcription factor (TF) binding site analysis that provided an efficient methodology for both clustering genes and investigating their coregulation. We used a combination of established methods ranging from multi-dimensional scaling (MDS) to the evaluation of Jaccard coefficients for TF sharing, to a formula for a global TF enrichment score, that cannot be found jointly at present on the well-known web server Gene Pattern . For our type of investigation, namely the correlation of a small number of gene expression levels across a human population, MDS is a very useful method. It is more flexible than principal component analysis (PCA) [6, 7], since it can be implemented for any type of correlation coefficient calculation, in particular for Kendall's which, because it is based on relative rank, provides the least constrained way of calculating correlations. Though hierarchical clustering [8–10], which we also give for completeness, can be derived for any correlation coefficient, it does not exhibit, because of its tree like structure, the rich two dimensional organization of a MDS distance scheme. Though MDS has been applied to large data sets , we use it here for a small set of 34 genes, which are further reduced in number through considerations of experimental procedures and GO category filtering. Our methodology is helpful in exploring hypotheses about gene coregulation.
Response Variation Dataset
We filtered out genes showing too low gene expression (gene names in blue with a crossed line in Table S2, copy number<2), and kept the genes with experimental variation smaller than the donor population variation (colored in red in the last column of Table S2) at some statistical level. The variance comparison was done by an F-test with H0: vs H1: . Of the original list of genes, we retained the 13 with copy number greater than 2 and with an uncorrected p-value smaller than 0.05 for the F-tests. The highest false discovery rate  for the 13 genes is 10.5%. Because the effective multiplicity of infection (MOI) defined as the virus/cell ratio may vary among the experiments, we also measured the expression of viral RNA. We examined the effect of adjustment of the gene copy numbers by dividing by NDV expression, which gave a larger list of 27 genes with p-value no greater than 0.05 for the F-test. However, some genes are induced only in infected cells whereas others are expressed similarly in both infected and uninfected cells. The aforementioned NDV correction procedure may introduce bias for the genes whose expression is not directly related to the MOI. The 13 genes that are selected from the data without NDV correction also appear significant in the data after NDV correction. To be conservative, we used the unadjusted 13 filtered genes for our subsequent analysis.
We hypothesized that genes showing similar patterns of change across infected individuals would be likely to share common genetic mechanisms responsible for these patterns. Our first goal was therefore to develop an approach to systematically identify and represent the correlation among the different genes across the population. We tested several metrics for pairwise correlation and represented the results using multidimensional scaling and, for completeness, hierarchical clustering. The approach is shown schematically in Figure 1B.
is minimized, where d i, jis the geometrical Euclidean distance between the representative points (xi, yi) and (xj, yj). Further details of the MDS algorithm can be found in reference . The geometric distance between any pair of genes reflects the dissimilarity in relative levels between the two genes among the 145 samples (Figure 4). The MDS approximation is relatively good, as given by the stress function 0.075, 0.092 and 0.17 for the Pearson's, Spearman's and Kendall's coefficients, respectively. The analysis for the three measures of dissimilarity showed the grouping of two clusters: (CCL5, IFNA1, IFNA2, IFNB1, IFIT2, IL29) and (IFIT1, IKBKE, IL6, IRF7, MX1, TBK1).
The MDS analysis can also be applied to the 145 samples and the 6 experimental repeats of a single sample (Figure S1; see Additional file 2) to look for correlation between individuals rather than correlation between genes. We can see there is some degree of within experimenter reproducibility, and it seems that when all genes are combined, between experimenter variation is not too different from donor variation. One of the experimental repeats (#4) appears to be an outlier. Therefore, we carried out an analysis both including and excluding this measurement and obtained comparable results.
Based on the MDS analysis of the 13 genes, the two gene clusters showed significant correlation (see Figure 4). We assessed the statistical significance of the two groups by generating randomized datasets that leave the distribution of each gene intact. For each randomization, the gene expression levels of the 145 samples of a gene were randomly permuted. An independent permutation was applied for each gene separately. For the measured MDS plot based on Kendall's correlation (Figure 4), the observed cluster dimensions, defined as the average of all pairwise distances among the genes, were 0.129 and 0.178 for the aforementioned two clusters. For the first cluster (CCL5, IFIT2, IFNA1, IFNA2, IFNB1, IL29), at each randomization, we recomputed the cluster dimension. Based on 4000 randomizations, we found that only one of the cluster dimensions of this six-member cluster was no greater than the observed dimension (0.129), i.e., the six genes were strongly correlated with high level of significance (p-value about 0.0003). We further evaluated the statistical significance of the observed cluster dimension. In each randomization, instead of computing the cluster dimension, we isolated the six-member cluster that gives the smallest cluster dimension. With the same 4000 randomizations, about four of the cluster dimensions of the best six-member clusters were no greater than the observed one, i.e. the group formed a tight cluster with a significant p-value (0.001). When applied to the second cluster (IFIT1, IKBKE, IL6, IRF7, MX1, TBK1), the same analysis showed that the six genes were highly correlated with a significant p-value of 0.0006 and formed a tight cluster with a significant p-value of 0.016.
Transcription Factor Sharing Analysis
Transcription Factor Enrichment Analysis
where s i and p i are the score and p-value computed for transcription factor i. In almost all cases, the condition p i ≤ 0.05 ensures a nonnegative score s i so TSSc takes into account all significant scores. A cluster's TSSc measures the extent to which genes in the cluster were co-regulated by transcription factors. For example, when evaluating a cluster of size 2 taken from the 12 genes (IL29 was removed because it lacks a TF entry), for which n 1 = 2 and n 2 = 10, a p-value no greater than 0.05 occurs only when k 1 = 2 and k 2 = 0, i.e. the TF should bind to all genes in the two-member cluster, but not to any gene outside of it. The results of the enrichment analysis are listed in Table S4A [see Additional file 2]. For GO term filtered data (see the TF sharing analysis section), the TF enrichment analysis is only based on 9 genes since IFIT1, IFIT2, IL29 and TBK1 have no TF entry, and the results are listed in Table S4B [see Additional file 2]. For the two clusters identified in the MDS plot, we are left with two four-member clusters (CCL5, IFNA1, IFNA2, IFNB1) and (IKBKE, IL6, IRF7, MX1) after we removed the aforementioned no TF entry genes. In Table S5 [see Additional file 2], among all 126 possible four-member clusters, the greatest TSSc of 4.8 belongs to (CCL5, IFNA1, IFNA2, IFNB1). The next greatest TSSc belongs to (IKBKE, IL6, IRF7, MX1). Thus the two clusters identified by MDS are most likely to be coregulated in terms of TSSc with a p-value of 1/126 = 0.008, and 2/126 = 0.016, respectively, which is consistent with the randomization results described earlier.
For the GO term filtered TF enrichment analysis, we checked the relation between expression distance as given by MDS and the total significance score TSSc. The correlation for all pairs is relatively weak (r = -0.329), which is probably due to many zero entries (Table S4B), in comparison with the correlation between MDS and TF sharing (r = -0.753). The correlation for the four-member clusters between the cluster dimension (defined as the average distance among all pairs of genes) and the TSSc is somewhat stronger (r = -0.456). This relatively weak correlation with MDS distance for the TF enrichment analysis is probably due to the fact that the genes in our analysis are biologically related, which tends to keep at low values the enrichment scores calculated for genes belonging to a cluster.
The TF enrichment analysis results, when compared with those of MDS clustering, need to be interpreted with caution, even when one ignores the presumed occurrence of many false positives. The analysis described above leads to two groups of 4 TFs each in which each of the two clusters is enriched relative to the other. These TFs however, despite the GO filtering by the main category of Immune system process (the others are Response to virus, and IFNA and IFNB receptor binding), do not appear to belong directly to the NDV immune response. Their appearance reflects the fact that across a wide spectrum of immune responses that includes cancer, the genes in the two clusters considered are effectively enriched, and in such a way that their total enrichment scores are larger than those associated with any cluster of four genes chosen from the set of 9 genes (see Table S5). To search for TF enrichment with a set of TFs with a presumed connection to the NDV immune response, we considered a list based on the transcriptional program in macrophages , and pathogenic  and common responses  (See Methods for details). This list includes 67 TFs and 98 motifs. An enrichment analysis gives IRF1 for cluster (CCL5, IFNA1, IFNA2, IFNB1), and CREM, E2F1, E2F6, E2F7, STAT1, STAT2 for cluster (IKBKE, IL6, IRF7, MX1). The results here are more attuned to the experimental situation at hand. In particular STAT1,2 are directly involved through ISGF3 in the transcription of IRF7 and MX1. For other expected TFs in the NDV immune response such as IRF3 or NF-kB (CCL5, IFNB1) that are part of the list, they bind to motifs in the promoter regions of genes in both clusters, have therefore very low enrichment scores, and do not show up in the final result. Thus enrichment analysis has limitations for interpreting the measured cluster structure. One reason is presumably that the relatively small number of genes central to the NDV immune response that are measured at 10 hours after infection are all more or less synchronously up-regulated, which makes it hard to distinguish between clusters unless enrichment encompasses categories of stimulation and cell types that go beyond the confines of the experiment proper. While our focus here was on TFs with known binding site preferences defined in TRANSFAC, future work could apply ab initio methods on these gene clusters to define novel TF candidate binding sites .
Web Tool for Coregulation Analysis
In order to facilitate the use of our methodology by other researchers, we have developed an interactive web-based correlation mapping tool called CorEx. It allows one to combine correlation maps in experimental data, based on hierarchical clustering and MDS analysis, with TF enrichment analysis. A user can choose to constrain the TF prediction by multiple phylogenetic conservations such as chimp, mouse or vertebrates. CorEx uploads the experimental data through the web interface, analyzes correlations and visualizes the hierarchical clustering and MDS. It also presorts genes according to the extent of the corresponding gene correlation. A user can then choose the interesting clusters for TF enrichment analysis. The application will find TFs having binding sites significantly enriched in the promoters of genes forming the submitted clusters. CorEx can be accessed at http://clip.med.yale.edu/corex/corex.php. In addition to the website, we implemented a web service that provides programmatic access to MDS, hierarchical clustering and TF enrichment analysis. This makes it easier to integrate these analyses in existing applications and provides a framework for coregulation-centered analysis of experimental data.
Our analysis showed that genes which MDS collects into one cluster are coregulated in the sense of sharing transcription factor binding sites, and are as well enriched in TF binding sites when compared with genes outside of the cluster. Our sample of measured genes is small at the start of our investigation, and further reduced by the consideration of experimental uncertainties, followed by GO categories filtering. These steps, designed to obtain statistically meaningful results, are unavoidable precautions that need to be taken all along the investigation. Even the list of TFs needed to study enrichment must be given careful consideration, as we discussed in the corresponding section when replacing the set of TFs in TRANSFAC by an immune response related set. This leads however to a new problem, the fact that some important immune response genes such as NF-kB that bind to many promoter regions, do not show up in the enrichment analysis because of their ubiquity. Our methodology thus highlights a number of issues that beset clustering analysis coupled with TF enrichment, whatever the number of genes investigated. Nevertheless, the overall computational approach developed herein, which is applicable to any type of multivariate data measured in individuals, and is accessible on our web server, provides an efficient hypothesis generator for the mechanisms underlying response variation in humans.
Differentiation of DCs
All human research protocols for this work have been exempted by the IRB of the Mount Sinai School of Medicine due to use of discarded samples not traceable to source. Monocyte-derived conventional DCs were obtained from buffy coats of human blood donors following a standard protocol . Briefly, human monocytes from buffy coats were isolated by Ficoll (Histopaque, Sigma Aldrich) density gradient centrifugation and CD14+ monocytes were immunomagnetically purified by using a MACS CD14 isolation kit (Miltenyi Biotech.). CD14+ monocytes (0.7 × 106 cells/ml) were later differentiated into immature cDCs after 5-6 day incubation in DC growth media [RPMI Medium 1640 (Gibco), 10% fetal calf serum (Hyclone), 2 mM of L-glutamine, 100 units/ml penicillin, 100 μg/ml streptomycin (Pen/Strep) (Invitrogen), 500 units/ml hGM-CSF (Preprotech) and 1000 units/ml hIL-4 (Preprotech)] at 37°C.
Virus preparation and viral infection
The recombinant Hitchner strain of Newcastle Disease Virus (rNDV/B1) was prepared and aliquots of allantoic fluid were harvested as previously described . NDV virus stock was titered by infection of Vero cell plates and identification of viral growth by the addition of monoclonal antibodies specific for NDV HN protein (Mount Sinai Hybridoma Core Facility) followed by addition of anti-mouse IgG-FITC and visualization using fluorescent microscopy. Titered NDV stock was diluted in Dulbecco's Modified Eagle Medium (DMEM) and added directly into pelleted DCs at a multiplicity of infection (MOI) of 0.5 prepared as previously described . After incubation for 30 minutes at 37°C, fresh DC growth medium was added back to the infected DCs (1.0 × 106 cells/ml). Virus free allantoic fluid was added to additional tubes of cells to serve as a negative control.
Ten hours after infection by NDV, DCs were pelleted and total RNA was extracted. The mRNAs of 34 genes and beta-actin were assayed by two step real time RT-PCR. The dried sample total RNA was resuspended into 12 μl DEPC water with 40 ng/μl dT (18 bp) primer. After heating for 10 min at 65°C and placing on ice for 2 min, each sample was mixed with 8 μl RT reaction mix (1 μl Stratscript (Stratagene, Texas) reverse transcriptase (50 U/μl), 2 μl 10× Stratscript buffer, 1 μl 10 mM dNTPs, 4 μl DEPC water) and heated for 2 hrs at 42°C then heated for 15 min at 70°C. The expression levels of 34 genes were determined by real-time PCR using the ABI Prism 9700 HT (Applied Biosystems). All real-time PCR assays were carried out in triplicate in a reaction containing: 1× Jumpstart Taq (Sigma-Aldrich Co.) buffer; 5 mM MgCl2; 0.2 mM each dNTPs (dUTP replacing dTTP); 0.5 μM each primers; 0.5 × SYBR Green (Invitrogen); 0.0125 U/μl Jumpstart Taq polymerase; cDNA template; final volume 10 μl. Cycling conditions for all genes were: 95.0°C for 2 min, followed by 40 cycles of 95.0°C for 15 sec, 55.0°C for 15 sec and 72.0°C for 30 sec. The PCR primer sequences are listed in Table S6 [see Additional file 2].
Hierarchical clustering and MDS
The mRNA copy numbers were collected from 145 samples described above, and were then transformed by the log2 function and further normalized by the subtracting the median. The matrix for each of the three types of correlations was computed using the statistical programming language R [26, 27] by the function cor. The detailed mathematical formulation is described in  or in the Supporting Information, Correlations Computations [see Additional file 2]. One defines the dissimilarity by one minus the correlation; the dissimilarity matrix can be used to construct the dendrogram for the hierarchical clustering by the R function hclust and the two dimensional coordinates of the genes for the multidimensional scaling (MDS) by the R function cmdscale. All R functions are run with version 2.8.0.
For predicting TFs binding to regulatory regions of the studied genes, we considered three gradually decreasing levels of phylogenetic conservation: (i) across rat, mouse, dog and human, (ii) across mouse and human, (iii) across chimp and human. Details of chimp and human conservation are as follows, with the others being similar: We used the aligned regions between human (version hg18) and chimp (version panTro2), which can also be downloaded from the UCSC genome bioinformatics website . The promoter region [-2k, 0] from the transcription starting site (TSS) is obtained for each human gene, where TSS is defined by the March 2006 refGene table . The TRANSFAC MATCH  with a cutoff chosen to minimize the sum of false positives and negatives is used to identify putative transcription factor binding sites for the human promoter region and the aligned chimp promoter region for all vertebrate transcription factor matrices in the 2009.1 release of TRANSFAC . The putative binding sites were considered to be phylogenetically conserved if matches were found in both human promoter regions and aligned chimp promoter regions of the same gene and no gaps were present in the human chimp alignment. Only three of the genes in this study (MX1, IRF7 and CREM) were associated with multiple TSSs. In these cases, we used the union of putative transcription factor binding sites from all of the alternate promoters. Vertebrate TRANSFAC matrices were included only if they could be linked to a HGNC  gene symbol, either directly or by an alias. A TF gene is listed as a presumed connection to the NDV immune response only if its gene symbol or its alias also appeared in Table S5 of reference , Table S8 of reference  or Table 1 of reference .
Jaccard coefficient for genes sharing TFs
Let S 1 and S 2 be the sets of transcription factors that bind to genes g1 and g2, respectively. The Jaccard coefficient for the two genes sharing TFs is , the ratio of the numbers of TFs in the intersection set of S 1 and S 2 to the number of TFs in the union set of S 1 and S 2.
The CorEx application consists of two interrelated elements: a CorEx analyzer and a front end web-interface. CorEx is scripted in PERL, calls R subroutines for the analysis, and presents the results via the Common Gateway Interface. The CorEx analyzer retrieves input data through the web-interface to generate MDS plots and a corresponding correlation matrix. The hierarchical clustering plot is generated as well if requested (see corresponding section of Methods). The analyzed result is displayed through the web-interface. The CorEx source code is available through the GNU General Public License and accessible at SourceForge upon request. The home page of CorEx is http://clip.med.yale.edu/corex/corex.php.
We thank Audrey Pendleton, Arnaud Ganee and Yaomei Ding for technical assistance. We thank Susanna Márquez for assistance in setting up the website and help in implementation of the web service and Robert Pfeffer for reading the manuscript. This work was supported by NIH NIAID Contract No. HHSN2662000500021C and Grant U19 AI62623. JLD was supported in part by NIH Grant T15 LM07056.
- Thomas DL, Thio CL, Martin MP, Qi Y, Ge D, O'Huigin C, Kidd J, Kidd K, Khakoo SI, Alexander G, et al.: Genetic variation in IL28B and spontaneous clearance of hepatitis C virus. Nature 2009,461(7265):798–801.View ArticlePubMedGoogle Scholar
- Goldstein DB, Cavalleri GL: Genomics: understanding human diversity. Nature 2005,437(7063):1241–1242.View ArticlePubMedGoogle Scholar
- Ernst J, Bar-Joseph Z: STEM: a tool for the analysis of short time series gene expression data. Bmc Bioinformatics 2006, 7:191.View ArticlePubMedGoogle Scholar
- Bar-Joseph Z: Analyzing time series gene expression data. Bioinformatics 2004,20(16):2493–2503.View ArticlePubMedGoogle Scholar
- Reich M, Liefeld T, Gould J, Lerner J, Tamayo P, Mesirov JP: GenePattern 2.0. Nature Genetics 2006,38(5):500–501.View ArticlePubMedGoogle Scholar
- Raychaudhuri S, Stuart JM, Altman RB: Principal components analysis to summarize microarray experiments: application to sporulation time series. Pac Symp Biocomput 2000, 455–466.Google Scholar
- Granucci F, Vizzardelli C, Pavelka N, Feau S, Persico M, Virzi E, Rescigno M, Moro G, Ricciardi-Castagnoli P: Inducible IL-2 production by dendritic cells revealed by global gene expression analysis. Nat Immunol 2001,2(9):882–888.View ArticlePubMedGoogle Scholar
- Eisen MB, Spellman PT, Brown PO, Botstein D: Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci USA 1998,95(25):14863–14868.View ArticlePubMedGoogle Scholar
- Blader IJ, Manger ID, Boothroyd JC: Microarray analysis reveals previously unknown changes in Toxoplasma gondii-infected human cells. J Biol Chem 2001,276(26):24223–24231.View ArticlePubMedGoogle Scholar
- Chen X, Cheung ST, So S, Fan ST, Barry C, Higgins J, Lai KM, Ji J, Dudoit S, Ng IO, et al.: Gene expression patterns in human liver cancers. Mol Biol Cell 2002,13(6):1929–1939.View ArticlePubMedGoogle Scholar
- Tzeng J, Lu HH, Li WH: Multidimensional scaling for large genomic data sets. Bmc Bioinformatics 2008, 9:179.View ArticlePubMedGoogle Scholar
- Viswanathan GA, Nudelman G, Patil S, Sealfon SC: BioPP: a tool for web-publication of biological networks. BMC Bioinformatics 2007, 8:168.View ArticlePubMedGoogle Scholar
- Viswanathan GA, Seto J, Patil S, Nudelman G, Sealfon SC: Getting started in biological pathway construction and analysis. PLoS Comput Biol 2008,4(2):e16.View ArticlePubMedGoogle Scholar
- Zaslavsky E, Hershberg U, Seto J, Pham AM, Marquez S, Duke JL, Wetmur JG, tenOever BR, Sealfon SC, Kleinstein SH: Antiviral response dictated by choreographed cascade of transcription factors. The Journal of Immunology 2010, in press.Google Scholar
- Benjamini Y, Hochberg Y: Controlling the False Discovery Rate - a Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society Series B-Methodological 1995,57(1):289–300.Google Scholar
- Press W, Flannery B, Teukolsky S, Vetterling W, (eds): Numerical recipes in C: The art of scientific computing. 2nd edition. Cambridge University Press; 1992.Google Scholar
- Cox C, Cox M, (eds): Multidimensional scaling. Chapman and Hall/CRC; 1994.Google Scholar
- Andersen J, VanScoy S, Cheng TF, Gomez D, Reich NC: IRF-3-dependent and augmented target genes during viral infection. Genes Immun 2008,9(2):168–175.View ArticlePubMedGoogle Scholar
- Ramsey SA, Klemm SL, Zak DE, Kennedy KA, Thorsson V, Li B, Gilchrist M, Gold ES, Johnson CD, Litvak V, et al.: Uncovering a macrophage transcriptional program by integrating evidence from motif scanning and expression dynamics. PLoS Comput Biol 2008,4(3):e1000021.View ArticlePubMedGoogle Scholar
- Jenner RG, Young RA: Insights into host responses against pathogens from transcriptional profiling. Nat Rev Microbiol 2005,3(4):281–294.View ArticlePubMedGoogle Scholar
- Huang Q, Liu D, Majewski P, Schulte LC, Korn JM, Young RA, Lander ES, Hacohen N: The plasticity of dendritic cell responses to pathogens and their components. Science 2001,294(5543):870–875.View ArticlePubMedGoogle Scholar
- Bailey TL, Williams N, Misleh C, Li WW: MEME: discovering and analyzing DNA and protein sequence motifs. Nucleic Acids Res 2006, (34 Web Server):W369–373.Google Scholar
- Hu J, Sealfon SC, Hayot F, Jayaprakash C, Kumar M, Pendleton AC, Ganee A, Fernandez-Sesma A, Moran TM, Wetmur JG: Chromosome-specific and noisy IFNB1 transcription in individual virus-infected human primary dendritic cells. Nucleic Acids Res 2007,35(15):5232–5241.View ArticlePubMedGoogle Scholar
- Park MS, Garcia-Sastre A, Cros JF, Basler CF, Palese P: Newcastle disease virus V protein is a determinant of host range restriction. J Virol 2003,77(17):9522–9532.View ArticlePubMedGoogle Scholar
- Hu J, Iyer-Biswas S, Sealfon SC, Wetmur J, Jayaprakash C, Hayot F: Power-laws in interferon-B mRNA distribution in virus-infected dendritic cells. Biophys J 2009,97(7):1984–1989.View ArticlePubMedGoogle Scholar
- Ihaka R, Gentleman R: R: A language for data analysis and graphics. Journal of Computational and Graphical Statistics 1996, 5:299–314.View ArticleGoogle Scholar
- R: A language and environment for statistical computing. Version 2.8.0 [http://www.r-project.org]
- Karolchik D, Baertsch R, Diekhans M, Furey TS, Hinrichs A, Lu YT, Roskin KM, Schwartz M, Sugnet CW, Thomas DJ, et al.: The UCSC Genome Browser Database. Nucleic Acids Res 2003,31(1):51–54.View ArticlePubMedGoogle Scholar
- Pruitt KD, Tatusova T, Maglott DR: NCBI Reference Sequence (RefSeq): a curated non-redundant sequence database of genomes, transcripts and proteins. Nucleic Acids Res 2005, (33 Database):D501–504.Google Scholar
- Kel AE, Gossling E, Reuter I, Cheremushkin E, Kel-Margoulis OV, Wingender E: MATCH: A tool for searching transcription factor binding sites in DNA sequences. Nucleic Acids Res 2003,31(13):3576–3579.View ArticlePubMedGoogle Scholar
- Matys V, Fricke E, Geffers R, Gossling E, Haubrock M, Hehl R, Hornischer K, Karas D, Kel AE, Kel-Margoulis OV, et al.: TRANSFAC: transcriptional regulation, from patterns to profiles. Nucleic Acids Res 2003,31(1):374–378.View ArticlePubMedGoogle Scholar
- Bruford EA, Lush MJ, Wright MW, Sneddon TP, Povey S, Birney E: The HGNC Database in 2008: a resource for the human genome. Nucleic Acids Res 2008, (36 Database):D445–448.Google Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.