Integrating genetic and network analysis to characterize genes related to mouse weight

Abstract

Systems biology approaches that are based on the genetics of gene expression have been fruitful in identifying genetic regulatory loci related to complex traits. We use microarray and genetic marker data from an F2 mouse intercross to examine the large-scale organization of the gene co-expression network in liver, and annotate several gene modules in terms of 22 physiological traits. We identify chromosomal loci (referred to as module quantitative trait loci, mQTL) that perturb the modules and describe a novel approach that integrates network properties with genetic marker information to model gene/trait relationships. Specifically, using the mQTL and the intramodular connectivity of a body weight-related module, we describe which factors determine the relationship between gene expression profiles and weight. Our approach results in the identification of genetic targets that influence gene modules (pathways) that are related to the clinical phenotypes of interest.

Figure 1. Visual Representations of the Gene Co-Expression Network

(A) Hierarchical clustering of 3,421 genes and visualization of gene module partitioning. The colored bars (below) directly correspond to the module (color) designation for the clusters of genes. One can visualize where in the whole clustering dendrogram the gene modules were defined. (B) Multi-dimensional scaling plot representation of the entire gene expression network. Each gene is represented by a dot, where the color of the dot corresponds to the gene module to which that gene belongs. The distance between each dot is representative of their topological overlap. This representation allows one to visualize the relationships among genes within a gene module, how that module is related to the rest of the network, and how closely any two modules are related.

Figure 2. The Relationship between the Blue Module and Several Clinical Traits

Module significance is defined as the mean of the absolute value of the correlation coefficient for all genes within a module with a physiological trait of interest. A module significance of 0.30 is statistically significant at a p-value of 0.05 for any module given n = 135 mice after correction for multiple comparisons (see Materials and Methods). Abfat, abdominal fat (g); Aneurysm, aneurysms as measured by histological examination using a semi-quantitative scoring method. AorticCal.L, aortic calcification in the lesion area; AorticCal.M, medial aortic calcification; Chol, total plasma cholesterol (mg/dl); FFA, plasma free fatty acids (mg/dl); HDL, high-density lipoprotein fraction of cholesterol (mg/dl); Index, adiposity index (total fat × 100/weight); Ins, plasma insulin (μg/l); LDL+VLDL, plasma low-density lipoprotein and very low-density lipoprotein choleseterol (mg/dl); Lesion, aortic lesion size as measured by histological examination using a semi-quantitative scoring method; MCP-1, plasma MCP-1 protein levels; otherfat, subcutaneous, retroperitoneal, and mesenteric fat (g); TG, plasma triglycerides (mg/dl); Totalfat, Abfat + otherfat; UC, plasma unesterified cholesterol (mg/dl).

Figure 3. Blue Module mQTL Profile Partially Overlaps with the Physiological Trait (Weight) QTL

The black curves represent the QTL analysis across the genome for mouse body weight using a single marker genome scan according to the LOD score scale to the right. Evidence of linkage for weight was found on Chromosome 1 (50.4 Mb, LOD = 3.03), Chromosome 5 (118.5 Mb, LOD = 3.39), Chromosome 15 (78.2 Mb, LOD = 3.58), and Chromosome 19 (50.4 Mb, LOD = 4.07). For each marker, the blue vertical line represents the number (divided by 20) of significant Blue module genes, i.e., the number of genes whose single-point LOD at this marker is bigger than 2. For example, there is a SNP marker on Chromosome 2 at which 117 Blue module genes have a LOD bigger than 2. The red stars denote the locations of the mQTLs reported in Table 1.

Figure 4. Relating the Gene Significance Measure for Weight (GSweight) to Genetic and Network Based Variables of the Blue Module

(A) Scatterplot between GSweight (y-axis) and intramodular connectivity (k_me) based on the module eigengene (x-axis). Each point corresponds to a gene in the Blue module. (B) Scatterplot between GSweight (y-axis) and the sum of the mQTL significance measures of Chromosomes 2, 5, and 10 (GSmQTL* = GSmQTL2 + GSmQTL5 + GSmQTL10a). (C) Scatterplot between GSweight (y-axis) and the mQTL significance measure of Chromosome 19 (GSmQTL19). (D) Scatterplot between GSweight (y-axis) and the predicted value based on a multivariable regression model involving intramodular connectivity (k_me) and the aforementioned mQTL significance measures (GSmQTL* and GSmQTL19). The linear regression model explains 70% of the variation in GSweight. (E) Boxplot for visualizing the effect of the module based variables on GSweight. The 8 different boxplots correspond to the eight groups of genes that result by splitting the module variables by their median values. Genes with high/low GSmQTL* values are labeled by q+ and q−, respectively. Genes with high/low GSmQTL19 values are labeled by 19+ and 19−, respectively. Genes with high/low connectivity values are labeled by k+ and k−, respectively (For example, the boxplot labeled q+ 19− k− plots the GSweight distribution of the Blue module genes with a high GSmQTL*, a low GSmQTL19, and a low connectivity).