A comparison of expressed sequence tags (ESTs) to human genomic sequences.

The Expressed Sequence Tag (EST) division of GenBank, dbEST, is a large repository of the data being generated by human genome sequencing centers. ESTs are short, single pass cDNA sequences generated from randomly selected library clones. The approximately 415 000 human ESTs represent a valuable, low priced, and easily accessible biological reagent. As many ESTs are derived from yet uncharacterized genes, dbEST is a prime starting point for the identification of novel mRNAs. Conversely, other genes are represented by hundreds of ESTs, a redundancy which may provide data about rare mRNA isoforms. Here we present an analysis of >1000 ESTs generated by the WashU-Merck EST project. These ESTs were collected by querying dbEST with the genomic sequences of 15 human genes. When we aligned the matching ESTs to the genomic sequences, we found that in one gene, 73% of the ESTs which derive from spliced or partially spliced transcripts either contain intron sequences or are spliced at previously unreported sites; other genes have lower percentages of such ESTs, and some have none. This finding suggests that ESTs could provide researchers with novel information about alternative splicing in certain genes. In a related analysis of pairs of ESTs which are reported to derive from a single gene, we found that as many as 26% of the pairs do not BOTH align with the sequence of the same gene. We suspect that some of these unusual ESTs result from artifacts in EST generation, and caution researchers that they may find such clones while analyzing sequences in dbEST.