Mapping gene expression quantitative trait loci by singular value decomposition and independent component analysis

Abstract

Background: The combination of gene expression profiling with linkage analysis has become a powerful paradigm for mapping gene expression quantitative trait loci (eQTL). To date, most studies have searched for eQTL by analyzing gene expression traits one at a time. As thousands of expression traits are typically analyzed, this can reduce power because of the need to correct for the number of hypothesis tests performed. In addition, gene expression traits exhibit a complex correlation structure, which is ignored when analyzing traits individually.

Results: To address these issues, we applied two different multivariate dimension reduction techniques, the Singular Value Decomposition (SVD) and Independent Component Analysis (ICA) to gene expression traits derived from a cross between two strains of Saccharomyces cerevisiae. Both methods decompose the data into a set of meta-traits, which are linear combinations of all the expression traits. The meta-traits were enriched for several Gene Ontology categories including metabolic pathways, stress response, RNA processing, ion transport, retro-transposition and telomeric maintenance. Genome-wide linkage analysis was performed on the top 20 meta-traits from both techniques. In total, 21 eQTL were found, of which 11 are novel. Interestingly, both cis and trans-linkages to the meta-traits were observed.

Conclusion: These results demonstrate that dimension reduction methods are a useful and complementary approach for probing the genetic architecture of gene expression variation.

Figure 1

Genome-wide distribution of linkages from single trait analyses. The yeast genome was divided into non-overlapping 20 kb bins and the number of significant linkages to that bin was recorded. In total, 5013 gene expression traits showed significant linkage (FDR = 0.05) in the single trait analyses. The probability of any bin having 20 linkages or more by chance is less than 2.1E-4. This is denoted by the solid red line (Methods). Details about each linkage hotspot are summarized in Table 1.

Figure 2

Proportion of Variance explained by each eigentrait. Singular value decomposition results in dimensions that we refer to as "eigentraits", which are ranked according to how much variation in the dataset they explain. The distribution of variance explained for the observed and null data are shown as blue and red circles, respectively.

Figure 3

Genome-wide linkage analysis for the top 20 eigentraits. In each linkage profile, the negative log p-value of the linkage statistic for each eigentrait is plotted against the genomic position of all the markers. Significance is determined by a GWER < 0.05. Fourteen of the 20 eigentraits show linkage to at least one QTL. Tolerance is set at 1E-10 for p-values equal to zero.

Figure 4

Genome-wide linkage analysis for the top 20 ICAtraits. In each linkage profile, the negative log p-value of the linkage statistic for each ICAtrait is plotted against the genomic position of all the markers. Significance is determined by a GWER < 0.05. Seventeen of the 20 ICAtraits show linkage to at least one QTL. Tolerance is set at 1E-10 for p-values equal to zero.

Figure 5

Overlap between results from single trait linkage scans with linkage analysis of eigentraits. The y-axis of the top panel represents the number of linkages that fall in each 20 kb non-overlapping regions spread across the genome while along the x-axis the position of genotyped markers at which linkage was estimated is marked. The next two rows of solid circles mark the position of eQTLs detected for eigentraits and ICAtraits, respectively. The position is aligned with the markers on the x-axis of the top plot. The red solid circles represents novel eQTLs while black represents previously described eQTL.

Figure 6

Overlap of QTL location across orthogonal eigentraits. In each plot the length of overlap between eQTL locations across multiple eigentraits is represented by the 1 LOD support interval that are denoted by black bars. Also, for each plot the location of the putative regulator, which might explain the enrichment of certain GO functional categories for traits correlated with specific eigentraits is represented along the x-axis by blue vertical bars.