Inferring details of past evolution from DNA is tricky
Last week I discussed one of many papers that use the ratio of protein-changing to "neutral" genetic changes, along the branches of an evolutionary tree, to infer past natural selection. This week's paper presents data calling that approach into question. This does not necessarily undermine the overall conclusions of last week's paper, which were based on a variety of methods, including testing the actual performance of mutant proteins.
"Hotspots of biased nuclear substitutions in human genes" was published in PLoS Biology by Jonas Berglund and colleagues. I am not a molecular biologist, so will just summarize their main points. The paper is open access.
Most of our DNA does not code for proteins. Some of the noncoding DNA is known to have important regulatory functions. But there is lots of DNA whose function, if any, is unknown, but which is nonetheless highly similar among species, as if any change was lethal. Except, when someone tried deleting this DNA, a bit at a time, most of the deletions were not lethal or even (as far as they could tell) harmful. I discussed this work earlier.
Anyway, much of this noncoding DNA that differs little among most species is different in humans. Could these differences be what makes us different from other apes? Quite possibly. But are all these human-vs.-chimp differences important? Maybe not. An unexpectedly high fraction of the changes from the ape ancestor we share with chimps involved a change from A bound to T (a weak bond) to G bound to C (a strong bond). Unless noncoding DNA with stronger bonds is consistently better somehow (and only in humans!), this suggests that these changes are caused by some DNA-specific process and not by natural selection. In other words, these changes occurred whether or not they were beneficial, just as mutations do. Could similar AT=>GC changes have changed protein-coding sections of DNA?
The researchers compared 10,238 genes in humans, chimps, and macaques...