Although sequencing of the human genome is nearly complete, deciphering its genes, estimated to represent less than 5% of the total sequence, is still a challenging problem. One popular strategy for identifying genes relies on expressed sequence tags (ESTs), which can be reassembled together to reconstruct the genome's coding regions. But this approach misses a significant fraction of genes, according to a report in November's Nature Genetics (26, 315–318, 2000). The study, by David Rank and co-workers, is based on a clever application of microarray technology to the task of gene discovery. The researchers began by designing PCR primers from predicted open reading frames in 350 megabases of genomic sequence. They amplified 9,498 500-base sequences directly from genomic DNA and spotted the amplicons onto microarrays. Hybridization with labeled cDNA from seven tissues and three cell lines showed that 51% of the arrayed sequences were expressed in at least one of the ten samples, and of these, about 15% were expressed in only a single sample. Many of the latter turned out to be novel (i.e., absent from GenBank's EST database). First author Sharron Penn says the technique “allows the confirmation of putative genes predicted by a series of gene finding algorithms.”