Deceptive combined effects of short allele dominance and stuttering: an example with Ixodes scapularis, the main vector of Lyme disease in the U.S.A.
Null alleles, short allele dominance (SAD), and stuttering increase the perceived relative inbreeding of individuals and subpopulations as measured by Wrights FIS and FST. Ascertainment bias, due to such amplifying problems are usually caused by inaccurate primer design (if developed from a different species or a distant population), poor DNA quality, low DNA concentration, or a combination of some or all these sources of inaccuracy. When combined, these issues can increase the correlation between polymorphism at concerned loci and, consequently, of linkage disequilibrium (LD) between those. In this note, we studied an original microsatellite data set generated by analyzing nine loci in Ixodes scapularis ticks from the eastern U.S.A. To detect null alleles and SAD we used correlation methods and variation measures. To detect stuttering, we evaluated heterozygote deficit between alleles displaying a single repeat difference. We demonstrated that an important proportion of loci affected by amplification problems (one with null alleles, two with SAD and three with stuttering) lead to highly significant heterozygote deficits (FIS=0.1, p-value<0.0001). This occurred together with an important proportion (22%) of pairs of loci in significant LD, two of which were still significant after a false discovery rate (FDR) correction, and some variation in the measurement of population subdivision across loci (Wrights FST). This suggested a strong Wahlund effect and/or selection at several loci. By finding small peaks corresponding to previously disregarded larger alleles in some homozygous profiles for loci with SAD and by pooling alleles close in size for loci with stuttering, we generated an amended dataset. Except for one locus with null alleles and another still displaying a modest SAD, the analyses of the corrected dataset revealed a significant excess of heterozygotes (FIS=-0.07 as expected in dioecious and strongly subdivided populations, with a more reasonable proportion (19%) of pairs of loci characterized by significant LD, none of which stayed significant after the FDR procedure. Strong subdivision was also confirmed by the standardized FST corrected for null alleles (FST=0.19) and small effective subpopulation sizes (Ne=7).