QTL/microarray approach using pathway information
© Matsuda et al; licensee BioMed Central Ltd. 2012
Received: 9 September 2011
Accepted: 15 January 2012
Published: 15 January 2012
A combined quantitative trait loci (QTL) and microarray-based approach is commonly used to find differentially expressed genes which are then identified based on the known function of a gene in the biological process governing the trait of interest. However, a low cutoff value in individual gene analyses may result in many genes with moderate but meaningful changes in expression being missed.
We modified a gene set analysis to identify intersection sets with significantly affected expression for which the changes in the individual gene sets are less significant. The gene expression profiles in liver tissues of four strains of mice from publicly available microarray sources were analyzed to detect trait-associated pathways using information on the QTL regions of blood concentrations of high density lipoproteins (HDL) cholesterol and insulin-like growth factor 1 (IGF-1). Several metabolic pathways related to HDL levels, including lipid metabolism, ABC transporters and cytochrome P450 pathways were detected for HDL QTL regions. Most of the pathways identified for the IGF-1 phenotype were signal transduction pathways associated with biological processes for IGF-1's regulation.
We have developed a method of identifying pathways associated with a quantitative trait using information on QTL. Our approach provides insights into genotype-phenotype relations at the level of biological pathways which may help to elucidate the genetic architecture underlying variation in phenotypic traits.
In the last decade, a large number of quantitative trait loci (QTL) have been mapped in various organisms, and positional information on QTL is already included in resources such as Online Mendelian Inheritance in Man , Mouse Genome Database  and AnimalQTLdb . Most previously reported QTL were detected by linkage analyses, but mapping resolutions were too low to identify underlying genes or causative mutations by direct experiments (i.e., the confidence intervals often span tens of map units that contain dozens to hundreds of genes). On the other hand, microarray-based gene expression profiling can be used to discover genes related to a quantitative trait of interest . Combining information on QTL with microarray data enables genes underlying genetically complex traits to be linked to QTL regions. Wayne and McIntyre (2002)  proposed the QTL/microarray approach to discover candidate genes. Typically, differentially expressed genes within a QTL region are identified, and candidates revealed based on the known function of a gene in the biological process governing the trait of interest. Because selections are based on limited pre-existing information on what processes may be involved, the nomination of candidate genes is biased . This simplistic approach focuses on gene discovery, which does not provide functional links between biological processes and traits. A low cutoff value in the individual gene analyses (IGAs) may result in many genes with moderate but meaningful changes in expression being missed, leading to a reduction in statistical power [7, 8].
Analyzing expression quantitative trait loci (eQTL) in specific tissues is an alternative way to identify potentially causative genes [9, 10]. eQTL derived from a polymorphism located near a gene (cis-eQTL) can be distinguished from distant eQTL (trans-eQTL). If cis-eQTL are collocating with phenotypic QTL and their expression is correlated with phenotype, they are regarded as strong candidates . A recent genome-wide association study (GWAS) found that trait-associated SNPs are significantly more likely to be eQTL, and annotating SNPs with eQTL scores can be useful for discovering susceptibility loci . However, the eQTL approach is restricted in its application due to the cost of genome-wide profiling for a large number of animals. In recent years, many methods for transcriptome analyses have shifted the focus from the IGA to a gene set analysis (GSA) to assess the differential expression of pre-defined gene sets that share common biological functions. The GSA has stronger statistical power than the IGA, and been shown to identify many novel gene sets with 'subtle but coordinated' expression patterns . Similar pathway-based approaches have been adopted in GWAS analyses, with some modifications to address the unique challenges of GWAS data [13, 14].
The present study aims to identify pathways associated with a quantitative trait of interest using information on QTL. We have developed a pathway-based QTL/microarray approach which provides candidate genes and biological insights into quantitative traits. Our approach is based on a modified GSA, that tests whether a subset of genes within a QTL region is more likely to be relevant to a trait of interest than whole pathway genes. We demonstrate that the combination of gene expression profiling, QTL mapping and GSA produced gene set scores suitable to detect trait-associated pathways.
Results and discussion
Phenotype strain survey data from Mouse Phenome Database
High-density lipoprotein cholesterol [mg/dL] age 7-9wks
Insulin-like growth factor 1 (adjusted serum IGF-1) [ng/mL] age 15-16wks
No. of mice
No. of mice
HDL cholesterol QTL set
The HDL cholesterol-related pathways detected
Directionof change in gene expression*
No. of genes
No. of genes within QTL regions
Drug metabolism - cytochrome P450
Environmental Information Processing
Arachidonic acid metabolism
Linoleic acid metabolism
Ascorbate and aldarate metabolism
Pentose phosphate pathway
Starch and sucrose metabolism
IGF-1 QTL set
The IGF-1-related pathways detected
Directionof change in gene expression*
No. of genes
No. of genes within QTL regions
All four comparisons
PPAR signaling pathway
Drug metabolism - other enzymes
Three comparisons without B6 vs NZB
ErbB signaling pathway
Environmental Information Processing
GnRH signaling pathway
MAPK signaling pathway
Environmental Information Processing
Steroid hormone biosynthesis
Gastric acid secretion
In this study, we have developed a method of identifying pathways that are associated with quantitative traits using information on QTL. In the application of our approach to the data on mouse inbred strains, the metabolic pathways for HDL cholesterol trait and the IGF-1 related signaling pathways were clearly detected. Our approach provides insights into genotype-phenotype relations at the level of biological pathways which may help to elucidate the genetic architecture underlying the variation in phenotypic traits. The method presented here is also useful for identifying positional candidate genes in the QTL region through the detection of trait-associated pathways. Once trait-associated pathways are detected, differentially expressed genes that share the pathways would be promising candidates, even if the genes are outside the QTL region.
Current QTL/microarray approach
With the current method, the simplest and most widely used fold change is calculated as a gene-level statistic by the log-ratio (base 2) of the group means of one gene.
In addition to an absence of transformation for the typical single direction enrichment, absolute values of per-gene test statistics were used to identify bi-directional changes, providing a more precise understanding of transcriptome regulation.
- 3)The most common choice for the gene set statistic is the mean of local statistics, especially when the gene-level scores are assumed to follow a normal distribution. We employed Z-scores for each gene set which were calculated from (absolute) fold change values for each gene between two experimental groups, as follows:
- 4)If a pathway is associated with the trait, we expect a sub-set of pathways within the QTL regions to have altered transcription. We intersected two gene sets, each of which belongs to the pathway and QTL category so that the intersection has composite information. Statistical tests conducted use a competitive null hypothesis. Suppose there are m gene sets (G i ) in the pathway category and these p-values are computed from the Z-score. For the given QTL category, the sorted p-values for sub-sets are
where iα/m means the critical value in the Benjamini-Hochberg procedure. In the current approach, an additional filtering based on the comparison of significance between sub- and whole pathways is implemented, although the method is not testing any specific null hypothesis for comparison of sub- with whole pathways. This approach assumes that the sub-set of pathways within the QTL region is under cis-regulation and partial trans-regulation within multiple QTL regions.
We applied our methods to two sets of mouse QTL identified from a public database. We extracted 47 and 17 QTL associated with blood concentrations of HDL cholesterol and IGF-1 in mice from MGI. For HDL cholesterol, we used annotations derived from the MP term "abnormal circulating HDL cholesterol level" (MP:0000184). For IGF-1, we used annotations from the term "abnormal circulating insulin-like growth factor I level" (MP:0004700). The QTL studies, the QTL name, peak position and confidence intervals were downloaded from MGI. If QTL flanking markers were not known, we assigned an interval by peak marker ± 20 Mb. To assign base pair and cytogenetic positions to all flanking and peak markers, we downloaded the UCSC Genome Browser  positional tables and parsed them corresponding GeneBank accession numbers. Base pair positions of the transcript first (5') and last (3') nucleotides are compared with the QTL base pair positions on the respective chromosome. If a gene overlapped with a QTL, information for the QTL is linked with the gene data.
Bennet et al. (2010)  compared gene expression profiles from 99 strains of inbred and recombinant inbred mice in their eQTL study. The data set came from the gene expression omnibus (GEO)  under accession no. GSE16780. We extracted data on C3H, B6, NZB, and SM strains including for liver tissue of 16-week-old male mice from the GSE16780 dataset. The platform for the data was the Affymetrix HT Mouse Genome 430A oligonucleotide microarray (22,716 probe sets). The data was processed using the Affymetrix GCOS algorithm utilizing the Robust Multi-array Average method (RMA) to determine the specific hybridization signal for each gene. We used expression values computed by merged triplicate arrays for the same strain with an average value. After filtering the Affymetrix control probe sets named starting with "AFFX-" which are generally used to evaluate hybridization performance and quality of the biological sample, to identify the single most appropriate probe set for each gene, the probe set with the maximum absolute expression value was uniformly assigned to the gene. In the end, we obtained 14,471 probe sets uniquely representing each gene/transcript covered by the Affymetrix HT Mouse Genome 430A array.
Each gene is subsequently annotated with its associated biological pathways, obtained from KEGG, a collection of manually drawn pathway maps representing the molecular interaction and reaction networks for 1) metabolism, 2) genetic information processing 3) environmental information processing, 4) cellular processes, 5) organismal systems and 6) human diseases. We limited the gene set size between ten and five hundreds and the resulting number of gene sets was 151 with 4,115 genes on the microarray (after removal of human diseases category maps). If QTL information is about eQTL regions, other gene sets such as transcription factor-target genes are needed to consider the characteristic feature.
List of abbreviations used
expression quantitative trait loci
false discovery rate
gene expression omnibus
gene set analysis
genome wide association studies
high density lipoproteins
individual gene analysis
insulin-like growth factor 1
Kyoto encyclopedia of genes and genomes
Mouse Genome Informatics
Mouse Phenome Database
quantitative trait locus
Robust Multi-array Average method.
- Hamosh A, Scott AF, Amberger JS, Bocchini CA, McKusick VA: Online Mendelian Inheritance in Man (OMIM), a knowledgebase of human genes and genetic disorders. Nucleic Acids Research. 2005, D514-517. 33 DatabaseView ArticleGoogle Scholar
- Blake JA, Bult CJ, Kadin JA, Richardson JE, Eppig JT, : The Mouse Genome Database (MGD): premier model organism resource for mammalian genomics and genetics. Nucleic Acids Res. 2011, D842-848. 39 DatabasePubMedPubMed CentralView ArticleGoogle Scholar
- Hu ZL, Fritz ER, Reecy JM: AnimalQTLdb: a livestock QTL database tool set for positional QTL information mining and beyond. Nucleic Acids Res. 2007, D604-609. 35 DatabaseGoogle Scholar
- Yano K, Imai K, Shimizu A, Hanashita T: A new method for gene discovery in large-scale microarray data. Nucleic Acids Res. 2006, 34: 1532-1539. 10.1093/nar/gkl058PubMedPubMed CentralView ArticleGoogle Scholar
- Wayne ML, McIntyre LM: Combining mapping and arraying: an approach to candidate gene identification. Proc Natl Acad Sci USA. 2002, 99: 14903-14906. 10.1073/pnas.222549199PubMedPubMed CentralView ArticleGoogle Scholar
- Fisher P, Hedeler C, Wolstencroft K, Hulme H, Noyes H, Kemp S, Stevens R, Brass A: A systematic strategy for large-scale analysis of genotype-phenotype correlations: identification of candidate genes involved in African trypanosomiasis. Nucleic Acids Res. 2007, 35: 5625-5633. 10.1093/nar/gkm623PubMedPubMed CentralView ArticleGoogle Scholar
- Mootha VK, Lindgren CM, Eriksson KF, Subramanian A, Sihag S, Lehar J, Puigserver P, Carlsson E, Ridderstråle M, Laurila E, Houstis N, Daly MJ, Patterson N, Mesirov JP, Golub TR, Tamayo P, Spiegelman B, Lander ES, Hirschhorn JN, Altshuler D, Groop LC: PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nat Genet. 2003, 34: 267-273. 10.1038/ng1180PubMedView ArticleGoogle Scholar
- Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, Paulovich A, Pomeroy SL, Golub TR, Lander ES, Mesirov JP: Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci USA. 2005, 102: 15545-15550. 10.1073/pnas.0506580102PubMedPubMed CentralView ArticleGoogle Scholar
- Schadt EE, Monks SA, Drake TA, Lusis AJ, Che N, Colinayo V, Ruff TG, Milligan SB, Lamb JR, Cavet G, Linsley PS, Mao M, Stoughton RB, Friend SH: Genetics of gene expression surveyed in maize, mouse and man. Nature. 2003, 422: 297-302. 10.1038/nature01434PubMedView ArticleGoogle Scholar
- Schadt EE, Molony C, Chudin E, Hao K, Yang X, Lum PY, Kasarskis A, Zhang B, Wang S, Suver C, Zhu J, Millstein J, Sieberts S, Lamb J, GuhaThakurta D, Derry J, Storey JD, Avila-Campillo I, Kruger MJ, Johnson JM, Rohl CA, van Nas A, Mehrabian M, Drake TA, Lusis AJ, Smith RC, Guengerich FP, Strom SC, Schuetz E, Rushmore TH, Ulrich R: Mapping the genetic architecture of gene expression in human liver. PLoS Biol. 2008, 6: e107- 10.1371/journal.pbio.0060107PubMedPubMed CentralView ArticleGoogle Scholar
- Doss S, Schadt EE, Drake TA, Lusis AJ: Cis-acting expression quantitative trait loci in mice. Genome Res. 2005, 15: 681-691. 10.1101/gr.3216905PubMedPubMed CentralView ArticleGoogle Scholar
- Nicolae DL, Gamazon E, Zhang W, Duan S, Dolan ME, Cox NJ: Trait-associated SNPs are more likely to be eQTLs: annotation to enhance discovery from GWAS. PLoS Genet. 2010, 6: e1000888- 10.1371/journal.pgen.1000888PubMedPubMed CentralView ArticleGoogle Scholar
- Wang K, Li M, Bucan M: Pathway-Based Approaches for Analysis of Genomewide Association Studies. Am J Hum Genet. 2007, 81: 1278-1283. 10.1086/522374PubMedPubMed CentralView ArticleGoogle Scholar
- Chasman DI: On the utility of gene set methods in genomewide association studies of quantitative traits. Genet Epidemiol. 2008, 32: 658-668. 10.1002/gepi.20334PubMedView ArticleGoogle Scholar
- The Mouse Phenome Database. http://www.jax.org/phenome
- Luoma PV, Sotaniemi EA, Pelkonen RO, Myllyla VV: Plasma high density lipoprotein cholesterol and hepatic cytochrome P-450 concentrations in epileptics undergoing anticonvulsant treatment. Scand J Clin Lab Invest. 1980, 40: 163-167. 10.3109/00365518009093020PubMedView ArticleGoogle Scholar
- Luoma PV, Sotaniemi EA, Pelkonen RO, Arranto A, Ehnholm C: Plasma high density lipoproteins and hepatic microsomal induction-relation to histological changes in the liver. Eur J Clin Pharmacol. 1982, 23: 275-282. 10.1007/BF00547568PubMedView ArticleGoogle Scholar
- The Mouse Genome Informatics Database.http://www.informatics.jax.org/
- Yusta B, Holland D, Koehler JA, Maziarz M, Estall JL, Higgins R, Drucker DJ: ErbB signaling is required for the proliferative actions of GLP-2 in the murine gut. Gastroenterology. 2009, 137: 986-96. 10.1053/j.gastro.2009.05.057PubMedView ArticleGoogle Scholar
- Murata Y, Konishi M, Itoh N: FGF21 as an Endocrine Regulator in Lipid Metabolism: From Molecular Evolution to Physiology and Pathophysiology. J Nutr Metab 2011. 2011, 981315-Google Scholar
- Zhen S, Zakaria M, Wolfe A, Radovick S: Regulation of gonadotropin-releasing hormone (GnRH) gene expression by insulin-like growth factor I in a cultured GnRH-expressing neuronal cell line. Mol Endocrinol. 1997, 11: 1145-1155. 10.1210/me.11.8.1145PubMedView ArticleGoogle Scholar
- Skrtic S, Wallenius K, Gressner AM, Jansson JO: Insulin-like growth factor signaling pathways in rat hepatic stellate cells: importance for deoxyribonucleic acid synthesis and hepatocyte growth factor production. Endocrinology. 1999, 140: 5729-5735. 10.1210/en.140.12.5729PubMedGoogle Scholar
- Kanehisa M, Goto S: KEGG: Kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000, 28: 27-30. 10.1093/nar/28.1.27PubMedPubMed CentralView ArticleGoogle Scholar
- Ackermann M, Strimmer K: A general modular framework for gene set enrichment analysis. BMC Bioinformatics. 2009, 10: 47- 10.1186/1471-2105-10-47PubMedPubMed CentralView ArticleGoogle Scholar
- Luo W, Friedman MS, Shedden K, Hankenson KD, Woolf PJ: GAGE: generally applicable gene set enrichment for pathway analysis. BMC Bioinformatics. 2009, 10: 161- 10.1186/1471-2105-10-161PubMedPubMed CentralView ArticleGoogle Scholar
- Kim SY, Volsky DJ: PAGE: parametric analysis of gene set enrichment. BMC Bioinformatics. 2005, 6: 144- 10.1186/1471-2105-6-144PubMedPubMed CentralView ArticleGoogle Scholar
- GAGE R package.http://sysbio.engin.umich.edu/~luow/downloads.php
- 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: 289-300.Google Scholar
- The UCSC Genome Browser.http://genome.ucsc.edu/
- Bennett BJ, Farber CR, Orozco L, Kang HM, Ghazalpour A, Siemers N, Neubauer M, Neuhaus I, Yordanova R, Guan B, Truong A, Yang WP, He A, Kayne P, Gargalovic P, Kirchgessner T, Pan C, Castellani LW, Kostem E, Furlotte N, Drake TA, Eskin E, Lusis AJ: A high-resolution association mapping panel for the dissection of complex traits in mice. Genome Res. 2010, 20: 281-90. 10.1101/gr.099234.109PubMedPubMed CentralView ArticleGoogle Scholar
- NCBI Gene Expression Omnibus. http://www.ncbi.nlm.nih.gov/geo/
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.