While there are many differences between tumor and non-tumor cells, the basic underlying distinction is in the DNA. Tumor cells harbor mutations, at least some of which are not present in non-tumor cells. Thus, a method of directly targeting cells containing specific mutations has potential for detection or treatment of cancer without the toxicity associated with more indirect approaches. Also, as mutations are a necessary component of malignancy, such a method is potentially applicable to all tumors. I propose a method by which several recently developed techniques can be utilized in a novel way to accomplish the goal of directly targeting mutations for cancer detection and therapy. The model can be summarized as follows: (1) Determine potential target mutations present in tumor cells but not non-tumor cells. (2) Construct molecules that will bind to DNA at the sites of mutation, but will not bind to DNA in normal cells. And, as a consequence of the molecules binding to the mutation, the cells will be destroyed. (3) Deliver these molecules to all cells (or at least all tumor cells). I hypothesize that such molecules can now be constructed using sequence-specific DNA binding proteins (such as customized zinc-finger DNA binding proteins) fused to transcriptional activator domains (such as VP16) and reporter or toxin genes. The necessary genes can be linked to the DNA binding proteins utilizing a recently described method based on expressed protein ligation.