A method of fragmentation of double stranded DNA is disclosed for use in nucleic acid analysis, notably in the multiplexed analysis of polymorphisms and mutations. The method produces a multiplicity of labeled sense and anti-sense fragments which are not complementary, and thus do not significantly re-anneal under conditions suitable for hybridization analysis (or capture-mediated elongation analysis) of the polymorphisms and/or mutations. The fragments display a desired or predicted length distribution. Cleavage sites can be selected such that the fragments are short, yet long enough to allow discrimination among fragments in an assay, and as a matter of statistical probability, such that the majority of fragments contain at least one labeled nucleotide to facilitate detection.