Semiautomated resolution of overlapping stutter patterns in genomic microsatellite analysis.

Microsatellites are polymorphic, short nucleotide repeating units scattered more or less randomly throughout the genome. They are readily detectable by polymerase chain reaction (PCR) and often used as genomic markers. One problem in the analysis of microsatellite data is the appearance of secondary bands during PCR that result in extended banding patterns. These "stutter" patterns may overlap in heterozygous alleles and obscure the overall pattern, severely interfering with analysis. This paper develops a model that successfully predicts the general shape of stutter patterns. It then presents techniques for measuring the intensity of the individual contributing alleles. The model is based on the assumption that there is a certain probability of losing or gaining a microsatellite repeat unit during each PCR cycle. The effect is cumulative, with the chance of losing a repeat unit being much greater than that of gaining one, which leads to a gradual reduction in the mean length of the pattern with increased PCR cycles. This can be modeled quantitatively to predict the shape of the stutter pattern, a prediction borne out by experiment. Next, a least-squares technique is presented that is used to analyze the overlapping stutter patterns and determine the relative concentration of each microsatellite in heterozygous alleles. The technique is based on the observation that, at least for microsatellites of approximately the same length, the relative intensity of each band in the stutter pattern is approximately the same for each allele. The stutter shape is most easily determined from homozygous alleles. It can also be approximated from heterozygous samples if the difference between the lengths of the primary microsatellite bands can be determined.