Is microsatellite mutational asymmetry detectable in allele frequency distributions?

We explore the mutational processes giving rise to microsatellite diversity by analysing allele lengths in 6045 human microsatellite markers drawn from the CEPH panel. Assuming a general mutation-drift process generating this diversity, the bias of the mutation distribution cannot be directly estimated from such data. However, inferences can still be made about the degree and sign of the asymmetry of the mutation distribution. We consider statistics based on moments of the observed length distribution, and derive their relevant analytical properties, showing that they have a high sampling variance. We conclude that moment estimators applied to allele length frequencies within the CEPH database could not be used to reject a null hypothesis of no bias even if bias was present. However, an order parameter does provide evidence of asymmetrically biased mutation: there is an unambiguous preponderance of alleles in which the shortest locus is the most frequent. It will be important to further characterize the sampling properties of such order parameters to estimate the magnitude of any mutation bias and the sensitivity of this estimation to the mutation model assumed.