Applied Biosystems approach works by chopping the genome up into 200 million pieces that are each about 1,500 DNA base pairs long, and then sequencing just the 25 base pairs at the edges of these pieces, yielding pairs of mated tags. By comparing the sequences of these DNA ends against a healthy reference genome as well as between the patients' normal and tumor genomes, the researchers could spot rearrangements between chunks of DNA. Velculescu's team found about nine regions of swapped DNA in each tumor, providing unique biomarkers for each patient's tumors.

The researchers then tracked the level of abnormal DNA as one of the colon cancer patients underwent different types of treatment. After surgery, chemotherapy, and surgical removal of metastases from the liver, the level of cancer-specific DNA in his blood dropped from 37 percent to 0.3 percent. This showed that some cancer cells still remained in the liver, indicating a need to remain vigilant and consider further treatment.

The research is an "exciting step down the road toward personalized cancer medicine," says Peter Johnson at the University of Southampton and Cancer Research UK's chief clinician. "The detection of DNA changes, unique to individual cancers, has proved to be a powerful tool in guiding the treatment of leukemia. If this can be done for other types of cancer like bowel, breast, and prostate, it will help us to bring new treatments to patients better and faster than ever."

Velculescu says the biggest caveat to a wide clinical use of the technique is the cost. While the price of sequencing has dropped dramatically, the analysis costs around $5,000 per genome. In an editorial accompanying the paper, Ludmila Prokunina-Olsson and Stephen Chanock of the National Cancer Institute, in Bethesda, MD, point out that researchers will need to sequence a number of cancer genomes before the approach can be put into clinical practice. For example, scientists need to assess how reliably these DNA rearrangements can be detected, whether certain types of rearrangements are most useful in tracking cancer, and whether certain parts of the genome tend to harbor these changes. In addition, researchers need to show that detecting latent cancer DNA can help tailor treatment, improving a patient's long-term health.