Viruses are always changing — forced to do so by the constant pressure to dodge the host immune system. RNA-based viruses, such as HIV and influenza, are infamous for evolving very rapidly, and so understanding the forces that shape viral evolution could have significant benefits for public health. One important question to consider is whether the rate of viral evolution is constant, and valuable insight into this question has recently been gained by two Danish researchers.

Dating where branches split on an evolutionary tree of viral sequences (or any sequence, for that matter) relies upon a constant rate of molecular evolution, known as the `molecular clock'. The intrinsic rate of the molecular clock, and the degree to which two sequences differ, allows one to infer the time at which the sequences diverged in the past. For viruses, this can help to pinpoint the origins of a particular epidemic.

However, statistical analyses of virus sequence evolution are inconsistent with the molecular clock model, and this has often been cited as evidence that the viral mutation rate changes in response to selection pressures. An alternative explanation — one that is tested by Schierup and Hein — is that the apparent variation in mutation rate is just an artefact caused by ignoring viral recombination.

The authors simulated, in silico, the evolution of 1,000-base-pair viral sequences. In each data run, the progenitor viral sequence was allowed to evolve up to an average divergence of 20%, but at a different recombination rate. The family of virus sequences that evolved in each run was used to build two phylogenetic trees — one with, the other without, assuming a molecular clock. In each experiment, the tree that is based on a molecular-clock model of evolution is taken as the null hypothesis. In this way, the two trees can be compared (using a likelihood ratio test) to measure the lowest amount of recombination at which the molecular clock is rejected (that is, the recombination rate above which the two trees are significantly different).

The results of the test are striking: as soon as the total number of recombinations in the history of ten sequences exceeds six (much fewer than actually occur), the molecular clock is rejected in more than 50% of runs. So viral sequences may well be evolving at a constant rate, but this isn't detectable because it is masked by recombination.

Can we still safely say, then, that the origin of the HIV1 pandemic can be dated to 1959, on the basis of a phylogenetic analysis? The pressure is on to invent ways of analysing viral sequence evolution in which the recombination rate is included as part of the method.