Serial analysis of gene expression (SAGE) is a powerful technique for the analysis of gene expression. A significant portion of SAGE tags, designated as orphan tags, however, cannot be reliably assigned to known transcripts. We used an improved reverse SAGE (rSAGE) strategy to convert human embryonic stem cell (hESC)-specific orphan SAGE tags into longer 3' cDNAs. We show that the systematic analysis of these 3' cDNAs permitted the discovery of hESC-specific novel transcripts and cis-natural antisense transcripts (cis-NATs) and improved the assignment of SAGE tags that resulted from splice variants, insertion/deletion, and single-nucleotide polymorphisms. More importantly, this is the first description of cis-NATs for several key pluripotency markers in hESCs and mouse embryonic stem cells, suggesting that the formation of short interfering RNA could be an important regulatory mechanism. A systematic large-scale analysis of the remaining orphan SAGE tags in the hESC SAGE libraries by rSAGE or other 3' cDNA extension strategies should unravel additional novel transcripts and cis-NATs that are specifically expressed in hESCs. Besides contributing to the complete catalog of human transcripts, many of them should prove to be a valuable resource for the elucidation of the molecular pathways involved in the self-renewal and lineage commitment of hESCs.