A mixed-oligonucleotide probe was used to identify four ras-like coding sequences in a human teratocarcinoma cDNA library. Two of these sequences resembled the rho genes, one was closely related to H-, K-, and N-ras, and one shared only the four sequence domains that define the ras gene superfamily. Homologs of the four genes were found in genomic DNA from a variety of mammals and from chicken. The genes were transcriptionally active in a range of human cell types.Mammalian ras genes (8,11,36) encode a family of proteins that show low but significant homology to the Go, subunits of G proteins (18). A number of genes encoding proteins with Mrs of 20,000 to 25,000 that share significant homology with the ras proteins have been isolated. The homology is greatest in four domains that have been shown through both mutagenic (see reference 2 for a review) and X-ray crystallographic (9, 17, 28) studies to be involved in the binding and hydrolysis of guanine nucleotides. Many ras-related proteins also contain a fifth conserved domain at their carboxy termini that, in H-, K-, and N-ras, is required for membrane localization and biological activity (13, 40). These ras-related proteins are found in a variety of eucaryotic organisms and appear to be well conserved over evolutionary time.The ras gene superfamily can be divided into several major groups on the basis of amino acid sequence: (i) the H-ras, K-ras, and N-ras proto-oncogenes (H, K, and N genes); (ii) the ral genes, which share about 50% homology with H-, K-, and N-ras (4, 5); (iii) the rap genes (29, 30) and R-ras (21), which differ significantly from each other but all share about 50 to 55% homology with the ras proteins, including strict conservation of the ras effector domain (amino acids 32 to 40 of H-ras); (iv) the rho genes, a more distantly related group that exhibits only about 35% identity with the ras proteins (23, 24); and (v)