Determination of allele frequency from DNA pools using bovine trinucleotide microsatellites

Three sets of fluorescent labelled primers were used to amplify bovine trinucleotide microsatellites from DNA pools. DNA from 20 individuals were collected to create 3 pools differing in allele frequencies. Replicate mixes from each pool were used as template for PCR reactions. PCR products were separated and quantified on an automated DNA sequencer. Allele frequency estimates from pooled samples corrected for overlapping shadow peaks were calculated. Rare alleles representing only 2.5% of the total pool were accurately detected. Standard error of allele frequency estimates expressed as percent of the total 40 chromosomes per pool ranged between 0.8%-4.6% for different microsatellite-pool combinations as compared to 8.0% binomial sampling error. Regression coefficients of actual allele frequencies, determined by individual genotyping, on estimated frequencies ranged from 0.96-1.06. As regression slopes were close to unity it can be deduced that corrected peak height values from a DNA pool are unbiased estimates of actual allele frequencies. With standard error of the y-intercept of 0.21, the 95% confidence interval of allele frequency is 0.42 alleles or 1% in a pool of 40 chromosomes. Thus, it would be possible to detect an allele with a frequency of greater than 1% within the pool.