Introduction

Insertion sequence (IS) is a kind of the mobile genetic elements. ISs are capable of independent DNA rearrangement and some of them are known to regulate their neighbor genes.Therefore, IS may contribute to the bacterial phenotype. Several insertion sequence families were identified based on relatedness of transposases, conservation of the catalytic site and organization, and the similar inverted repeats (IR). In this study, the purpose is to identify the ISs in bacterial whole genomes. Based on the typical IS element features, the blastx and inverted repeats finding are used without the current knowledge about IS families and IRs. The annotation is still included as a reference to indicate the result quality. Therefore, the hypothetical proteins may be included and some degenerated/truncated IS elements could also be predicted.

Figure 1, The structure of a typical IS element.

In this analysis, the IS finding is based on the structure of the typical IS element (fig. 1). The IS elements contains the direct repeats (DR) which is the target site duplication, left and right inverted repeats (IRL and IRR) and the open reading frames which may encode transposase. The features used in this analysis are IRs and the open reading frames which probably encode transposase. The size of IRs is about 10~40 bps and the total length of the IS element is less than about 2.5 kbps. The size of DRs are too small to be a feature, only 2~14 bps.

Figure 2, The flow chart before the mannually curation

		A.	Perform genomic DNA sequences blastx against the ISPEP database
		B.	Extend 1000bps at each end of the hit regions
		C.	Use Inverted Repeat Finder (IRF) to find inverted repeats
		D.	Attach gene annotations and calculate the D1 and D2 distance
			Filter the records with D1 or D2 larger than 100bps
			Filter the records with at least one IR region located in open reading frames
		E.	Perform the second blastx search of regions within IRs and delete the following cases:
			1.	Hit regions locate in the IR regions
			2. 	No blastx hits
			3.	The redundant hits with smaller IRF score
	

Material and Methods

1. Actinobacillus actinomycetemcomitans HK1651
2. Fusobacterium nucleatum ATCC 25586
3. Prevotella intermedia 17
4. Streptococcus mutans UA159
5. Streptococcus sanguinis SK36
6. Streptococcus gordonii Challis substr. CH1
7. Porphyromonas gingivalis W83
8. Tannerella forsythensis ATCC 43037
9. Treponema denticola ATCC 35405
10. Actinomyces naeslundii MG1

There are IS elements recorded in the ISfinder database. This database contains the DNA sequences or portein sequences in the IS elements and these IS elements are classified as families. The protein sequences (ISPEP) are searched by whole genome sequences with blastx. In order to get the most possible results, the cut-off is not set here because some truncated or degenerated transposases will not get good matches against the ISPEP. In this step, the initial hit regions (all the high scoring pairs (HSP)) are found to be candidated regions.

Tow of the most significant features of IS elements are the transposase coding genes inside and the inverted repeats at both terminal regions. The number of the open reading frames (ORFs) can be more than one. Since the boundries of the IS elements are not known at this step and the IS elements are about 2.5kbps, the initial hit regions are extended both sides 1kbps. Then the sequences of the initial hit regions with their flanking regions are used as input for IRF. This program considers some repeats features like the percents of matches and indels, GC content and so on. Then based on these features, the IRF calculate its score as measuring the goodness of the IRs. Here, the threshold of this score is lower than the IRF authors' recomendation because in some cases, the IRs may be degenerated through evolution and become not significant features. However, the basic intact structures of IS element are what we want. The initial hit regions with the IRs found are considered as containing the possible IS elements which are inside between the IRL and IRR.

The flanking region is 1000 base pairs each side so the region within IRs may not contain the transposases, which are the ISPEP hits. The second blastx against ISPEP database is to examine the regions within IRs. In this step, the regions within IRs with blastx hits against ISPEP are the possible IS elements. The evalue cut-off is set as 0.0001 because we know the boundries of the predicted IS elements. Therefore, the predicted elements contain the IR at terminal ends and at least one blastx ISPEP hit within IRs.

Even the predicted IS contains at least one blastx ISPEP hit, it could be random hits or fusion proteins. The regions within IRs found are also used to perform blastx against nr protein database to check the second ISPEP hits. The regions without any transposase hit will be deleted from our candidate IS elements. However, the results of the blastx are considered as the reference for checking the second ISPEP hits. No cut-off is set (evalue is 10) and all the results are preserved and the number of hits with "transposase", "hypothetical" and others are recorded. So, even if the results with only one non-significant hits against nr database will be preserved for further investigation.

No current annotations are used as the criteria for finding transposase. The annotations are considered as a reference for checking the result quality. In addition, the annotation also gives the information of the ORFs. If the annotation exists in the records, the second ISPEP blastx hits will be checked if the regions of hits locate in the ORFs. Only the hit regions locate in the ORFs are recorded as the predicted IS element records.

The purpose of refinement is to eleminate the redundant records. For example, the same IRs will be found in the slightly different regions of initial hits. The region with better evalue is preserved and the worse one is deleted. And if the two predicted IS element contains the same ORFs, the IR with higher IRF score is recorded.

After the refinement, there are still many records in nearly the same regions but with different sets of ORFs or with overlapped hits of the second ISPEP blsatx results. Therefore, before the mannually curation, those regions have to be combined as the final refinement results. Then each region is defined as the overlapped hits on the target genomes of the second ISPEP blastx results. That is, if the second ISPEP blastx hits on the genome are overlapped with other hits, these overlapped records are combined in one regions which starts from the smallest hit start to the largest hit end. And the records are defined as the records of predicted IS element which contains the terminal IR. In our result below, first you have to select the interested regions and then it shows the detailed records in this regions.

The mannually curation step is relatively simple in each regions. If the distances between IRs and the ORF are smaller, the higher chances this record is true. All the distances below 100 base pairs are kept and others are deleted in each region. If there are no such records, then all results are preserved. The other thing to curate is to check if the annotation is the transposase. If the annotations indicate other functions, the blastx results against ISPEP and nr databases are examined to see if this is a not significant match or a fusion protein.

For the no annotation parts, if the region contains both the records with and without annotation. The no annotation records are deleted. If all the records are without annotation in one region, all records are preserved. However, if these records are within ORFs, they should be deleted because they are partial matches with those ORFs. It is interesting to investigate the regions with all no annotation record are located in intergenic regions. This may be the results of wrong gene prediction or the results of the degeneration of IS elements, especially when the record gets long hits against ISPEP.

Data presentation

The user interface contains four parts, which are organism selection, regions, records and sequence/Blastx information.

1. Organism selection: You can select the interested organism here.
2. Regions: As the materials and methods descibe, one region is defined as the region contains overlapped hits on the target genomes in the second ISPEP blastx results. Once the organism is selected, the regions in this organisms are shown here.
3. Records: The records are the predicted IS elements with IRs at terminus. After each region is clicked, the records in this region shows here.
4. Sequences/Blastx information: If you click the column "IR locations", the sequences will be retreived and show in this field. If you click "the name of ISPEP blastx results" or "Blastx against nr database", both the graphical and text output results will show here.

1. This page is written in content-centric ajax framework. Except the links in column "ORF ids", which will open another pages to show gene annoation, all other links will not go to other pages.
2. You can toggle the content of last three parts by clicking the header with the gray background.

1. ISPEP blastx results
• Family: The family of the IS elements of the ISPEP hits
• Name: The name of the ISPEP hits
• Evalue and % identity: The evalue and % identity of blastx results
• Coverage: The coverage defined here is the blastx matched region length divided by the subject length (target length in ISPEP).
2. Annotations
• ORF ids: The ORF ids which are inside the predicted IS elements
• Descriptions: The descriptions of the above ORFs
• D1, D2: The length of D1 and D2 regions. D1 and D2 regions are the distance between IRL, IRR and ORF regions. Please see figure 1. If both of these lengths are smaller than 100 base pairs, the fonts of the whole row are bold.
3. Blastx against nr database: The summary of the blastx results against nr protein database (no cut-off).