The GeneZoo™ Or The GeneJungle™?

Over 40% of Stratagene’s GeneConnection™ cDNA gene clusters are novel— and more than 50% are estimated to be full-length clones

The GeneZoo™ or the GeneJungle™?

Jason Goncalves • Jean-Michel Lelias • Joe Sorge
Stratagene

The GeneConnection™ discovery cDNA clone collection^YY  goes beyond normal cDNA collections. Stratagene has searched through many different human tissues and has identified thousands of cDNAs that are not found in the UniGene database (our GeneJungle™).* We have systematically sequenced the ends of our clones and have used a rigorous cDNA clustering algorithm to determine how many of our clones are unique and how many can be found in UniGene (those found in UniGene are called our GeneZoo™). About 40% of our clusters are in the GeneJungle and 60% are in the GeneZoo. The average insert size of our cDNAs is 1.7 Kb, which indicates that about half of the clones are predicted to have full length open reading frames. We have taken extreme care to guarantee clone identity and purity. Stratagene’s clones can be found on our GeneConnection Virtual Lab using several different free search strategies, including searching by gene name, keyword, UniGene ID, or nucleotide or protein sequence (www.stratagene.com). A gel image of a restriction digest of each clone can be viewed on our website. Stratagene’s GeneConnection discovery clone collection has been tracked and curated with precision and consistency in mind from its inception, and the GeneJungle collection provides an untapped resource for new discoveries.

The raw sequence of the human genome is now mostly complete, but it will be quite some time before all of the genes and their exon structures are known. Gene finding software programs are inaccurate at spotting exon-intron junctions. They can split groups of contiguous exons into two putative genes when they really belong to one gene. They can also group exons into one gene when they really belong to two or more genes. Thus the importance of high quality cDNA collections is rising once again. cDNAs are an excellent way of confirming gene structure, and Stratagene has a highly unique, well-characterized collection of 1.7 Kb average length human cDNA clones. Approximately 40% of Stratagene’s cDNA clusters cannot be found in UniGene, and based on our initial in-house analysis of known sequences about 50% of Stratagene’s clones contain full-length open reading frames which corresponds to our predicted estimate for the collection.

Inherent Problems with Clustering ESTs

ESTs (Expressed Sequence Tags) are DNA sequences derived from sequencing the 5¢ and/or 3¢ termini of cDNA inserts using vector-specific sequencing primers. Each EST sequence is typically 300 to 600 nucleotides in length. The dbEST database currently contains about 2.1 million human ESTs. UniGene is an attempt to reduce these sequences to a non-redundant set of gene-oriented “sequence clusters”. In theory there should be one cluster for each underlying gene. Build #116 of UniGene (July 3, 2000) has produced 81,967 unique clusters. However UniGene is plagued with many artifacts. The ESTs found in dbEST have come from many different sources, with widely varying degrees of sequence quality. The quality of the cDNA libraries from which these ESTs were derived is variable. Sometimes two cDNAs can be fused together in one vector. This creates UniGene artifacts whereby two unrelated cDNAs are placed into the same UniGene cluster. Genomic DNA contamination is common, as well as are products resulting from aberrant transcription and termination. Splice variants are of course common, and it is difficult for the UniGene clustering algorithms to differentiate splice variants from different gene products.

Table 1
Common UniGene Artifacts and Stratagene’s Solutions

Stratagene has made some very unique cDNA libraries. Much has been done to minimize artifacts such as fusion inserts and genomic DNA inserts (Table 1). Moreover the cDNAs have been highly normalized, yielding a very low level of clone redundancy. We have systematically sequenced the 3¢ ends of our clones and analyzed the resulting sequences. In the analysis, we remove sequences lacking poly-A tails. This effectively eliminates an artifact typically found in UniGene. If 5¢ and 3¢ sequences from the same gene are non-overlapping in UniGene, then UniGene will put these 5¢ and 3¢ sequences into two different clusters, overestimating the number of human genes. Because we only cluster 3¢ end sequences bearing poly-A tails we do not see this 5¢/3¢ splitting artifact (Table 1). Moreover, because the sequences that we have clustered to date are contiguous with the poly-A tails, we are less likely to be confused by splice variants, which occur less frequently in the 3¢ untranslated regions of transcripts (Table 1). Utilizing the 3¢ sequences as a unique gene-identification tag has been demonstrated as an effective gene-specific marker. Because the 3¢UTRs are not as conserved as the coding sequences, this makes it easier to distinguish between individual genes and paralogous gene family members that may have sequence homology in their coding sequences¹.

Fig.2

Our clustering algorithm is rigorous. We first identify commonly repeated sequence elements in the data set. We then require that for any two sequences to cluster, they must match at 96% identity over 100 or more base pairs and the percentage of alignable sequence must be greater than 90%. Figure 2 shows how the percentage of alignable sequence is determined. Two sequences are compared and aligned to maximize the number of matching base pairs. At each end of the local alignment, the shorter of the two unaligned sequences is used to calculate the number of alignable bases. The number of alignable bases is simply the sum of the local alignment length plus the length of unaligned sequence flanking the local alignment. Of course, alignments of commonly repeated or low complexity sequences are discarded. The algorithm will not cluster sequences from different gene family members, since the untranslated regions tend not to align. In contrast, sequencing artifacts are ignored since they generally do not drop the percent identity below 96%. Our algorithm would place splice variants into different clusters; however since we only use the sequence contiguous to the poly-A tail to perform the clustering, splicing is not a significant factor. We could choose to ignore splice variants when clustering (Figure 2) by eliminating internally unpaired sequence from the computation of alignable length. However the algorithm we have chosen is more rigorous, and we rely instead on there being little splicing near the 3¢ ends.

Fig.1

We have also clustered our sequences together with 1.7 million human EST sequences that are included in the human UniGene Database (Build #116). Those clusters that contained a UniGene representative (Stratagene GeneZoo) were also compared with 41,472 sequence-verified IMAGE clones from Research Genetics, and 9,182 Unigem 2.0 clones from Incyte Genomics. Figure 1 shows that most clusters, except for those in Stratagene’s GeneJungle, and except for 354 Incyte clones, fall within the 81,967 UniGene clusters. Interestingly, Incyte’s 9,182 Unigem clones collapse into 8,298 UniGene clusters when referenced against UniGene build #116, plus 354 non-UniGene clusters. Research Genetics’ 41,472 IMAGE clones collapse into 31,521 unique clusters when referenced against UniGene build #116. (Table 2)

Table 2
Relative Overlaps of Various Human cDNA Clone Collections

Clustering was based on UniGene build #116 for sequences that match UniGene. For the 354 Incyte Unigem 2.0 clones that lie outside of UniGene it is assumed that each of the 354 represents an individual cluster.

Table 2 shows that Stratagene’s GeneConnection 1.0 Discovery set has a substantial proportion of clusters not found in UniGene (about 41% of the Stratagene clusters are in the GeneJungle). This suggests that Stratagene’s libraries contain rare sequences not commonly found in other cDNA libraries. With an average insert size of 1.7 Kb, the Discovery clone set contains over 50% full length human cDNAs. This suggests that out of 25,321 clones we currently have over 12,500 full-length sequence-tagged cDNAs, and over 5000 of these full-length cDNAs have never been reported publicly. Stratagene is expanding its collections and updating its website with additional sequences on a regular basis.

Searching for Stratagene Clones

To find clones within Stratagene’s collection, the GeneConnection website (www.stratagene.com) allows searches by keyword gene name, accession number, Unigene number, or DNA or protein sequence. Stratagene has annotated all clones to optimize searches using key words, so that a clone similar to a characterized gene can also be found, for example “ESTs, Highly similar to protein-tyrosine-phosphatase” or “Zinc finger protein homologous to Zfp-36 in mouse (ZFP36)”.

Fig.4

Searches can be carried out with nucleic acid or protein search queries. Search results will show the degree of match as both a % identity of the aligned bases and as the quality score of the match (Figure 4). Similar genes can be found this way. For example if you want to find homologs to a gene of interest, obtain the sequence of your gene of interest and paste it into the search window on the GeneConnection search page (see Figure 3). Several search engines are available. Nucleotide sequence target data can be searched with a nucleotide sequence query using simple BLASTN. Nucleotide sequence target data can be first translated into all 6 open reading frames and then searched with a query DNA sequence that is also translated into all 6 reading frames using the search engine TBLASTX. Nucleotide sequence target data can be translated into all 6 reading frames and searched with a query protein sequence using TBLASTN. If a match is found to a Stratagene clone, the clone information will be displayed. The Stratagene clones have all been restriction mapped and size estimates are available for all clones. Sample restriction gel data are available on the website for all clones. If the DNA sequence of the Stratagene clone is within the UniGene set (a GeneZoo clone) its DNA sequence will be revealed in the search report. If the DNA sequence of the Stratagene clone is outside of UniGene (a GeneJungle clone), its DNA sequence is provided upon standard purchase of the clone.

Fig.3

Since Stratagene has only entered 3¢ sequence information into its clustering database, we have designed an automatic “indirect match” strategy to help locate clones having homologous sequences in the UniGene database. Even if you enter a coding sequence or a 5¢ sequence, matches can be found through a bridging database. We have taken our 3¢ sequences and BLASTed them against all the sequences in the UniGene database. When identities above a certain threshold are found, the UniGene sequence is placed into an “indirect database” with a link to the homologous Stratagene clone. When you perform sequence searches at our GeneConnection website, the program automatically searches both our direct sequence data and the indirect database sequence data. Both types of matches are shown in separate sections of the search report. Indirect match reports show the alignment between your query sequence and the indirect (UniGene) sequence, with a link to the Stratagene clone name and number.

All of our GeneConnection discovery clones can be purchased as a bacterial stab culture. We use the XL10-Gold^® strain, which is T1-phage resistant thus minimizing the threat of T1-phage contamination. The clones in our collection are categorized as either GeneZoo clones or GeneJungle clones and are differentially priced accordingly. Refer to the website for price information and special discount prices are available for large volume orders.

Conclusions

While other clone collections may provide a defined subset of human cDNAs, Stratagene’s GeneJungle goes beyond the familiar territory of UniGene. If you have a desire to discover new genes or gene families, the GeneJungle is an exciting place to explore. With an average insert size of 1.7 Kb, the probability of finding a previously undiscovered, full length human cDNA is high since we estimate that half of our clones contain full-length open reading frames. All clones have been restriction mapped, so you can obtain an estimate of the insert size before ordering a clone. The clones have been carefully sequenced and tracked, assuring that you receive the clone you ordered.

REFERENCE

1. Wilcox, AS, et al. (1991) Nucleic Acids Research 19(8): 1837-1843.

* Patents pending