Most studies of ras oncogene activation use assays for ras mutations based on the polymerase chain reaction (PCR) of DNA segments containing ras exons 1 and 2, followed by allele-specific oligonucleotide (ASO) hybridization or direct sequencing, which require that to be detectable, a mutation must be present in at least 3-25% of ras alleles. Thus, studies of tissues in which only a fraction of cells contains a ras mutation risk false negative results. To minimize this risk, we have developed a highly sensitive, non-radioactive assay for ras mutations. Ras genes were PCR-amplified using mismatched primers, to introduce restriction sites into products derived from normal alleles. Repeated restriction digestion and PCR enriched for mutant alleles, visualized by agarose gel electrophoresis. Serially diluted DNA samples containing ras mutations demonstrated detection of 1 mutant/10(6) normal alleles (four orders of magnitude more sensitive than PCR/ASO hybridization). This assay was applied to DNA from four patients with relapsed acute leukemia in whom ras mutations present at diagnosis were not detectable by PCR/ASO hybridization at relapse. In one case, the mutation present at diagnosis was demonstrated at relapse. In the others, loss of the mutation was confirmed, at a greatly increased sensitivity. This method is widely applicable to detection of mutant ras alleles admixed with larger numbers of normal alleles.