Aging mice show a decreasing correlation of gene expression within genetic modules

Abstract

In this work we present a method for the differential analysis of gene co-expression networks and apply this method to look for large-scale transcriptional changes in aging. We derived synonymous gene co-expression networks from AGEMAP expression data for 16-month-old and 24-month-old mice. We identified a number of functional gene groups that change co-expression with age. Among these changing groups we found a trend towards declining correlation with age. In particular, we identified a modular (as opposed to uniform) decline in general correlation with age. We identified potential transcriptional mechanisms that may aid in modular correlation decline. We found that computationally identified targets of the NF-KappaB transcription factor decrease expression correlation with age. Finally, we found that genes that are prone to declining co-expression tend to be co-located on the chromosome. Our results conclude that there is a modular decline in co-expression with age in mice. They also indicate that factors relating to both chromosome domains and specific transcription factors may contribute to the decline.

Figure 1. A novel approach for differential co-expression network analysis.

(A) Collections of microarrays can be used to look for differences in co-expression networks with age. (B) Each edge in the difference network represents the change in correlation that occurs between young and old mice. The edge weights in the 16-month and the 24-month networks are a function of the correlation in expression between two genes. The edge weights in the difference network represent the magnitude of the difference in correlations. The correlation difference matrix view is a heat map representation of the difference network. In both the network view and the matrix view, red represents a decrease in correlation, and blue represents an increase.

Figure 2. Functional gene clusters tend to decrease with age.

(A) The number of co-expression- and Gene Ontology (GO)- defined groups that change with age. The red bars indicate decreasing co-expression, and the blue bars indicate increasing co-expression. The dotted line represents the number of groups expected under the null distribution. (B) Histogram of the sum of the edge weights for the clusters using and . Again the red bars indicate decreasing co-expression, and the blue bars indicate increasing co-expression. Top left panel: The total number of groups that decrease in co-expression exceeds any of the permuted values. Top right panel: The total number of groups that decrease in co-expression exceeds all but 4/1,000 of the permuted values.

Figure 3. Difference in the clustering coefficient () versus the connectivity () distributions between young and old networks.

Each dot represents a probe in either the 16-month-old (blue) and 24-month-old (red) networks. All of the probes with at least one neighbor are plotted.

Figure 4. Deletion simulations indicate modular co-expression loss.

(A) The clustering coefficient () versus the connectivity () after simulation of uniform co-expression loss. Each dot represents a probe in either the 16-month-old (blue) and 24-month-old (red) networks. All of the probes with at least one neighbor are plotted. The gray lines represent the 100 node-deletion simulations. (B) The clustering coefficient () versus the connectivity () for young and old mice as contrasted with the cluster-based deletion simulation. Each dot represents a probe in either the 16-month-old (blue) and 24-month-old (red) networks. All of the probes with at least one neighbor are plotted. The distributions from the cluster-based deletion simulations are shown in gray.

Figure 5. Heat map representation of correlation changes in targets of transcription factors that significantly decrease correlation with age.

Red represents a decrease in correlation between two genes, and blue represents an increase. The yellow boxes identify the subsets of genes that are strongly decorrelated.

Figure 6. Genes that are prone to correlation loss are clustered on the chromosome.

(A) For a window of a set size, genes with connectivity in the difference network above a threshold are counted. (B) The positions of genes are plotted on the chromosome. The black bars are genes that do not meet the loss score threshold. Genes with a high loss score (red bars) were found to be clustered together.