We have characterized two abundant human cDNAs which, through Northern hybridization analysis, are selectively expressed in human sarcomeric muscle. DNA sequencing was performed and the two cDNAs were found to share sequence identity, with the exception of a 3' UTR extension present on the longer transcript. Our data suggest that the two transcripts are generated through alternative use of two poly(A) addition signals. The cDNAs encode a large open reading frame encompassing at least 435 codons. Through sequence comparisons, both at the DNA and predicted amino acid sequence level, we have been unable to find significant sequence similarity to any other characterized sequence. Consequently, we have termed this novel human sequence sarcosin. Although novel, Southern hybridization analysis demonstrated that the sarcosin sequence has been conserved in several mammalian species.