Figure 3. The dependence of clustering count 
on sequence divergence.
A) The number of additional amino acid substitutions expected in the vicinity of a focal substitution due to clustering, 
, increases linearly with divergence between species. This is seen in this plot of 
against the the fraction of substituted amino acids 
for six comparisons of D. melanogaster to sister Drosophilids. B) Non-synonymous substitutions cluster more strongly in more constrained genes. Here 
is estimated from subsets of the aligned coding sequences for the species comparison of D. melanogaster and D. yakuba. The subsets corresponds to the ten deciles of the coding sequences ranked by non-synonymous divergence. More constrained genes (lower 
, darker color) have more pronounced clustering, seen as the larger peak of 
near 
. The inset shows clustering count 
versus the average 
of each subset. 
increases linearly at low 
, but quickly levels off and is roughly constant at 
for 
. This contrasts with the result in panel A where 
measured divergence time, rather than constraint, and 
increased linearly with 
.