Asymmetric sequence divergence of duplicate genes

Abstract

Much like humans, gene duplicates may be created equal, but they do not stay that way for long. For four completely sequenced genomes we show that 20%-30% of duplicate gene pairs show asymmetric evolution in the amino acid sequence of their protein products. That is, one of the duplicates evolves much faster than the other. The greater this asymmetry, the greater the ratio Ka/Ks of amino acid substitutions (Ka) to silent substitutions (Ks) in a gene pair. This indicates that most asymmetric divergence may be caused by relaxed selective constraints on one of the duplicates. However, we also find some candidate duplicates where positive (directional) selection of beneficial mutations (Ka/Ks > 1) may play a role in asymmetric divergence. Our analysis rests on a codon-based model of molecular evolution that allows a test for asymmetric divergence in Ka. The method is also more sensitive in detecting positive selection (Ka/Ks > 1) than models relying only on pairwise gene comparisons.

Figure 1

Schematic representation of our model tree. Two duplicates are presumed to have diverged from an outgroup gene. Each of the three branches of this tree is allowed to have its own K_s and K_a values.

Figure 2

Significance of observed differences in K_a. On the X-axis is plotted the absolute value of the difference in K_a value between two duplicates. The Y-axis gives the negative logarithm (to base 10) of the P value for that pair. (A) All asymmetric duplicate pairs shown. (B) Only the square region marked in panel A is shown. The dashed line in B shows the significance level of P = 0.05.

Figure 3

Correlation of the normalized difference in K_a between two duplicates and the difference in the selective constraint for those two duplicates. The variables on the axes labels are |ΔK_a| = |(K_a1 – K_a2)/(K_a1 + K_a2)| and |ΔK_a/K_s| = |(K_a1/K_s1 – K_a2/K_s2)/(K_a1/K_s1 + K_a2/K_s2)|, respectively. (A) S. cerevisiae. (B) D. melanogaster. (C) C. elegans.

Figure 4

Triplet-based analysis is more sensitive in detecting positive selection. Number of cases where at least one duplicate in a pair has K_a/K_s > 1.0 for the four organisms shown using our triplet method (black) and using the conventional pairwise estimation (grey). The total number of duplicates pairs for which we determined K_a/K_s in this analysis was 22 for S. cerevisiae, 14 for S. pombe, 44 for D. melanogaster, and 164 for C. elegans.