Single-nucleotide polymorphism-gene intermixed networking reveals co-linkers connected to multiple gene expression phenotypes
© Gong et al; licensee BioMed Central Ltd. 2007
Published: 18 December 2007
Gene expression profiles and single-nucleotide polymorphism (SNP) profiles are modern data for genetic analysis. It is possible to use the two types of information to analyze the relationships among genes by some genetical genomics approaches. In this study, gene expression profiles were used as expression traits. And relationships among the genes, which were co-linked to a common SNP(s), were identified by integrating the two types of information. Further research on the co-expressions among the co-linked genes was carried out after the gene-SNP relationships were established using the Haseman-Elston sib-pair regression. The results showed that the co-expressions among the co-linked genes were significantly higher if the number of connections between the genes and a SNP(s) was more than six. Then, the genes were interconnected via one or more SNP co-linkers to construct a gene-SNP intermixed network. The genes sharing more SNPs tended to have a stronger correlation. Finally, a gene-gene network was constructed with their intensities of relationships (the number of SNP co-linkers shared) as the weights for the edges.
It is increasingly recognized through genomic studies that genes regulated via the same mechanisms are likely to have similar mRNA expression profiles, for example, by sharing the same transcriptional factors, or pathways, or any unknown but important factor. Several investigators have provided indirect evidence for this hypothesis by clustering genes according to their mRNA expression profiles . Thus, it may be a feasible strategy to search for co-regulated gene clusters from the correlated genes linked to a common SNP(s) as the shared factor. The relationship between the co-expression and the shared factor(s) (which we call co-linkers) has not been directly tested or quantified on a large scale previously because it is difficult to provide a reliable estimate to measure such relationships using a small number of genes and factors. In this study, we were very interested in identifying such genetic factors (for instance SNPs). A SNP co-linker cis- or trans-linked with a number of genes may indicate that it is either a functional polymorphism or close to an underlying genetic co-factor nearby that is able to modulate these genes. Today, both genome-wide gene expression and SNPs can be measured at the same time, which allows identification of such cis- and trans-acting loci, often called (pleiotropic) expression quantitative trait loci (eQTLs) at the "omics" scale. By treating gene expressions as quantitative traits and SNPs as genomic landmarks, the analysis can proceed in the same (or extended) manner as mapping genetic loci for physiologic or clinical traits .
In this study, following the conventional linkage analysis for identification of the "susceptibility" SNP/loci for each gene, we further identified these SNP co-linkers by measuring the strength of the relationship between the linked genes and the co-linker. After finding SNP co-linkers, the SNP-gene intermixed network could be further expanded by linking two genes if they shared a common SNP co-linker. We used Haseman-Elston sib-pair linkage analysis to establish gene-SNP linkages. Then, hub genes and SNPs were identified by their high degree of connectivity. Finally, a SNP-gene intermixed network was constructed.
In this study, we used data for Problem 1 from Genetic Analysis Workshop 15 (GAW15), which provided expression levels of 3554 genes in lymphoblastoid cells from fourteen three-generation Centre d'Etude du Polymorphisme Humain (CEPH) Utah families. Genotypes of 2882 autosomal and X-linked SNPs were also included . For the sib-pair linkage analysis, allele frequencies of each SNP locus were estimated by using a maximum-likelihood method incorporated in the FREQ program of S.A.G.E. (Version 5.3) , and the identical-by-decent (IBD) data were produced by the GENIBD program in S.A.G.E. . Due to the problem for multipoint IBD computations of very dense SNPs, we performed only single-point IBD computations.
Sib-pair linkage analysis
where the summands were the gene expression values measured on N sib pairs. Then the dependent variable for the jth pair was
The basic regression model we fitted was of the form:
where α was the intercept, and β h and h were the total genetic variance due to the hth SNP marker and the estimated IBD at the marker respectively, and ε was the residual error. This regression model was used to establish the gene-SNP linkages.
Constructing a gene-SNP intermixed network
Exploring the expression pattern between genes linked to a common SNP locus
where r ij was the Pearson's correlation coefficient between ith and jth genes. A gene could be clustered into different functional groups if it was linked to multiple SNP co-linkers. A permutation approach was used to determine the consensus threshold for G i . We randomly sampled n genes from the original whole gene pool and calculated the averaged correlation coefficient between the n genes. The same process was repeated 100,000 times and the 95% quantile was then defined as the empirical threshold that corresponds to the type I error of 0.05 .
Extracting a gene-gene network
Gene expressions are often influenced by multiple factors (e.g., transcription factors, reporters, enhancers, modifiers, and so on). The more factors two genes share, the closer the relationship they tend to have. However, the strong linkage disequilibrium (LD) among the SNP co-linkers might influence the result. To investigate the issue, we assessed the LD for all SNP pairs by implementing a likelihood ratio test using the software Arlequin , whose empirical distribution was obtained by a permutation procedure . We found that LD had a small influence on the result (only 4.8% of gene pairs were co-linked with a SNP pair or block of a significant LD (p < 0.05)). As shown in Figure 1b, we defined two genes to be connected if they shared at least one linked SNP and the number of co-linked non-significant LD SNPs for the two genes was used to describe the intensity of their relationship (considering a SNP pair of a significant LD as a single SNP). A gene-gene network was then constructed according to the newly identified relationships with their intensities of relationship as the weights for the edges.
We found a total of 4727 gene-SNP linkage pairs at the significance level of 10-7 after multiple test correction using Bonferroni's method. There were 60 pairs defined as cis-acting where the SNPs localized within the extended 10-Mb region of the paired genes, and all the remaining were trans-acting.
Gene-SNP intermixed network and hubs
The top 1% genes of a high degree of connectivity
The top 1% SNPs of a high degree of connectivity
Co-expression pattern between genes linked to a common SNP locus
In our study, first we used a univariate method to identify gene-SNP relationships for each gene expression phenotype separately. After finding the pool of all gene-SNP relationships, we constructed the SNP-gene intermixed network and further gene-gene network based on the concept of sharing a factor(s). Alternatively, one may consider using purely multivariate modelling to include all the genes, but this is not practical yet due to the prohibitive computing demand. One may also consider several data reduction techniques such as principal-component analysis as seen in several applications in this workshop. From both theoretical and applied views, all the methods have their own merits and weaknesses in analyzing modern high-dimension data. Further comparisons between different approaches are expected to provide novel methodological insights.
Although we identified gene-SNP relationships using very strict criteria, it is likely that the resulting gene-SNP intermixed network (or the further derived gene-gene network) includes some noise links. Hub SNPs or hub genes, which are expected to play key roles in maintaining the underlying biological networks in a working condition, should be further studied either computationally or using molecular biology techniques prior to applying these hub SNPs or hub genes in more practical settings.
The proposed bioinformatics approach, which attempts to identify co-regulated gene clusters from the correlated genes linked to a common SNP(s) as the shared factor, has been demonstrated via analysis of the GAW15 Problem 1 data, to be a promising and feasible strategy to identify transcriptional QTLs (hub SNPs) and to elucidate the underlying transcriptional genetic networks on the global scale. However, further studies, including topological and functional evaluation and validation of the resulted networks using more data sets and large-scale simulations are warranted.
This work was supported in part by the National Natural Science Foundation of China (Grants 30370798, 30571034, 30570424, 30600367, and 30670484) and the Foundation of Heilongjiang Province (Grants 1055HG009, ZJG0501, GB03C602-4, F2004-02, 2005-39, and BMFH060044). Some of the results of this paper were obtained by using the program package S.A.G.E., which is supported by a U.S. Public Health Service Resource Grant (RR03655) from the National Center for Research Resources.
This article has been published as part of BMC Proceedings Volume 1 Supplement 1, 2007: Genetic Analysis Workshop 15: Gene Expression Analysis and Approaches to Detecting Multiple Functional Loci. The full contents of the supplement are available online at http://www.biomedcentral.com/1753-6561/1?issue=S1.
- Inoue LY, Neira M, Nelson C, Gleave M, Etzioni R: Cluster-based network model for time-course gene expression data. Biostatistics. 2007Google Scholar
- Rao S, Wang QK: SAGE programs: model-free linkage analysis for complex cardiovascular phenotypes. Methods Mol Med. 2006, 128: 61-89.View ArticlePubMedGoogle Scholar
- Cheung VG, Conlin LK, Weber TM, Arcaro M, Jen KY, Morley M, Spielman RS: Natural variation in human gene expression assessed in lymphoblastoid cells. Nat Genet. 2003, 33: 422-425. 10.1038/ng1094.View ArticlePubMedGoogle Scholar
- S.A.G.E.: Statistical Analysis for Genetic Epidemiology, Release 5.3. 2006, Cleveland: Department of Epidemiology and Biostatistics, Case Western Reserve University, [http://genepi.cwru.edu/]
- Li X, Rao S, Jiang W, Li C, Xiao Y, Guo Z, Zhang Q, Wang L, Du L, Li J, Li L, Zhang TW, Wang QK: Discovery of time-delayed gene regulatory networks based on temporal gene expression profiling. BMC Bioinformatics. 2006, 7: 26-10.1186/1471-2105-7-26.View ArticlePubMed CentralPubMedGoogle Scholar
- Schneider S, Roessli D, Excoffier L: Arlequin: A software for population genetics data analysis. Ver 2.000. 2000, Geneva: Genetics and Biometry Laboratory, Department of Anthropology, University of Geneva, [http://lgb.unige.ch/arlequin/]Google Scholar
- Slatkin M, Excoffier L: Testing for linkage disequilibrium in genotypic data using the expectation-maximization algorithm. Heredity. 1996, 76: 377-383. 10.1038/hdy.1996.55.View ArticlePubMedGoogle 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.