Cancers are often associated with mutations in multiple genes; thus, studying the distributions of genes that harbor cancer-promoting mutations in cancer samples and their co-occurrences could provide insights into cancer diagnostics and treatment. Using data from the Catalogue of Somatic Mutations in Cancer (COSMIC), we found that mutated genes in cancer samples followed a power-law distribution. For instance, a few genes were mutated in a large number of samples (designated as high-frequent genes), while a large number of genes were only mutated in a few samples. This power-law distribution can be found in samples of all cancer types as well as individual cancers. In samples where two or more mutated genes are found, the high-frequent genes, i.e., those that were frequently mutated, often did not co-occur with other genes, while the other genes often tended to co-occur. Co-occurrences of mutated genes were often unique to a certain cancer; however, some co-occurrences were shared by multiple cancer types. Our results revealed distinct patterns of high-frequent genes and those that were less-frequently mutated in the cancer samples in co-occurring and anti-co-occurring networks. Our results indicated that distinct treatment strategies should be adopted for cancer patients with known high-frequent gene mutations and those without. The latter might be better treated with a combination of drugs targeting multiple genes. Our results also suggested that possible cross-cancer treatments, i.e., the use of the same drug combinations, may treat cancers of different histological origins. (C) 2015 Elsevier Inc. All rights reserved.