We have investigated the capability of PCR to amplify a specific locus from each template in a mixture of allelic DNAs. Modeling experiments employed a 300-bp HOX2B segment as target and utilized, as distinct template "alleles," genomic DNAs from 1 human and from 2 chimpanzees known to differ in sequence at the target. Two modes were used: (1) mixtures of PCR products that had been previously amplified individually from a single template and (2) PCR amplification en masse from composite human/chimpanzee genomic DNA templates. Products generated by either mode were separated by denaturing gradient gel electrophoresis (DGGE). Detection of a "trace" allele mixed with a "dominant" one was possible by simple ethidium bromide staining of the gel up to a sensitivity of 1 part in 20. A balanced mixture was represented by 5:5 and 4:3:3 mixtures of allelic PCR products or genomic templates; an uneven mixture of dominant and trace alleles, by 9:1 and 8:1:1 mixtures. For 5:5, two homoduplex bands and two heteroduplex bands of equal intensity were generated. For 9:1, the trace homoduplex disappears whereas the two heteroduplexes are easily visible. For 8:1:1, four heteroduplexes and one homoduplex were observed; homoduplexes and heteroduplexes formed from the trace alleles were not visible. These experiments demonstrate that PCR can amplify mixed allelic templates in direct proportion to the stoichiometric fraction of each template. Trace species are captured as heteroduplexes with the most abundant species and are clearly displaced on denaturing gradient gels from the dominant homoduplex species. Our analytical studies can be applied to analysis of sequence variants in DNA collected from cancerous or infected tissues.(ABSTRACT TRUNCATED AT 250 WORDS)