Detecting the presence and diversity of low-level mutations in human tumors undergoing genomic instability is desirable due to their potential prognostic value and their putative influence on the ability of tumors to resist drug treatment and/or metastasize. However, direct measurement of these genetic alterations in surgical samples has been elusive, because technical hurdles make mutation discovery impractical at low-mutation frequency levels (<10(-2)). Here, we describe inverse PCR-based amplified restriction fragment length polymorphism (iFLP), a new technology that combines inverse PCR, RFLP, and denaturing high-performance liquid chromatography to allow scanning of the genome at several thousand positions per experiment for low-level point mutations. Using iFLP, widespread, low-level mutations at mutation frequency 10(-2)-10(-4) were discovered in genes located on different chromosomes, e.g., OGG1, MSH2, PTEN, beta-catenin, Bcl-2, P21, ATK3, and Braf, in human colon cancer cells that harbor mismatch repair deficiency whereas mismatch repair-proficient cells were mutation free. Application of iFLP to the screening of sporadic colon cancer surgical specimens demonstrated widespread low-level mutations in seven out of 10 samples, but not in their normal tissue counterparts, and predicted the presence of millions of diverse, low-incidence mutations in tumors. Unique low-level mutational signatures were identified for each colon cancer cell line and tumor specimen. iFLP allows the high-throughput discovery and tracing of mutational signatures in human cells, precancerous lesions, and primary or metastatic tumors and the assessment of the number and heterogeneity of low-level mutations in surgical samples.