We study properties of networks constructed from gene-expression data obtained from many types of carcinomas. The networks are constructed by connecting vertices that belong to each others list of K-nearest-neighbors, with K being an a priori selected non-negative integer. We introduce an order parameter for characterizing the homogeneity of the networks. On minimizing the order parameter with respect to K, degree distribution of the networks shows power-law behavior in the tails with an exponent of unity. Analysis of the eigenvalue spectrum of these networks confirms the presence of the power-law and small-world behavior. This shows that these gene networks contain a few very highly connected genes and a large number of genes with low connectivity. These results conclusively show that various carcinomas are a consequence of malfunction of a few genes that either regulate the expression of a large number of other genes or make the hubs of crowded regulatory pathways.