The time needed to update UniGene with new sequences varies. Generally, this takes more than 1 week but less than 1 month.

UniGene clusters can eventually be "retired" for various reasons, such as:

• the sequences in a cluster might be retracted by the submitters because they are found to have contaminants.
• two clusters can be joined, in which case one of the original cluster IDs would be retired.
• a cluster can be split into two or more clusters in such a way that none of the smaller clusters can be recognized to be "the same as" the original cluster.

Cluster IDs are not reused after being retired, and specific information about why a particular cluster was retired is not available from the UniGene Web pages.

Using a retired cluster number (Hs.######) in UniGene's search tool will generate a page with links to the current clusters for the sequences.

UniGene clusters often have an expression such as "ESTs, highly similar to ACTIN 1" or "weakly similar to..." How are the degrees of similarity defined?

   1. "Highly similar to" means >90% in the aligned region.

   2. "Moderately Similar to" means 70-90% similar in the aligned region.

   3. "Weakly similar to" means <70% similar in the aligned region.

 records calculated?

For each nucleotide sequence in UniGene, a search is made for sequence similarity to known proteins from eight organisms. This is done using Blastx. Blastx compares the six-frame conceptual translation products of a nucleotide query sequence (both strands) against a protein sequence database. Blastx has 'in-frame' gapped alignments and uses sum statistics to link alignments from different frames.

The peptide databases used by UniGene are those representing Homo sapiens, Mus musculus, Rattus norvegicus, Drosophila melanogaster, Caenorhabditis elegans, Saccharomyces cerevisiae, Escherichia coli. and Arabidopsis thaliana. The eight protein databases exclude mitochondrial proteins and are screened for redundancy.

The nucleotide sequence is considered to match the protein sequence if the BLASTx E value is less than 1e-6. For each of the eight databases, the best-hit is that alignment with the lowest bit score. This protein is used as one of the eight prot_sim features of the nucleotide sequence.

The proteins assigned to a UniGene cluster are chosen among the prot_sim proteins assigned to the cluster's component sequences. The exact algorithm used to select the representative protein is currently being revised.

Is there an easy way to get the best contig/alignment for all the EST data in a cluster?

Although no assembly or contig is available, the longest sequence in each cluster has been identified. Automatic assembly of EST sequences is periodically re-evaluated. To date, we have found that accurate assemblies require curation and have chosen not to create the inaccurate dataset.

There is a file on the NCBI anonymous FTP site called Hs.seq.uniq.Z. This contains one sequence selected from each UniGene cluster that is the one with the longest region of high-quality sequence data.

splice forms, all other sequences in the cluster will be subsequences of this one. You can find this file at ftp://ftp.ncbi.nih.gov/repository/UniGene/.

For example, a user interested in cluster Hs.10920 might search the file using a text editor or the command "grep" on UNIX for the string "Hs.10920". This procedure would identify the section of the file that contains the best sequence.

Can you give me more details on the construction of a particular library, such as the tissue source, cloning strategy, or vector used?

Information concerning library origin is only provided by the submitter of the sequences, and NCBI does not have any additional information other than what is shown in the record. You can contact the submitter using the contact information found in the dbEST records or GenBank records for ESTs derived from the library of interest.

How are the cluster titles assigned?

There are several possible sources for the title, in order of preference:

• For Human, LocusLink title. MGD data are used in an analogous way for mouse
• name of product
• defline of mRNA record
• defline of genomic record
• ESTs, similar to something
• ESTs

The mRNA or genomic sequence chosen is arbitrary from the end-user's perspective; there is no easy way for the user to look at the sequences in a cluster and reproduce the algorithm that chooses which one gives the title.

How do you define a polyadenylation signal in the UniGene sequences?

Some sequences have an obvious polyA tail. However, we allow only a finite amount of sequence beyond the end of the polyA tail. For this reason, tails on low-quality reads are sometimes missed. Other sequences are also considered as ending with a polyA tail if there is a polyA signal. The signal is a sequence of ATTAAA or AATAAA that is 10-35 nucleotides from either the polyA tail or sequence end. All sequences are searched in both orientations.

How can I link my page to UniGene?

Several possibilities exist for making links to UniGene's pages.

Each requires the organism of interest be specified, using the following abbreviations:

UniGene References