The Institute for Systems Biology Interspersed Repeat Protein Masking Documentation

Transposable element protein database

The transposable element protein database now contains 5411 predicted proteins, with a combined length of 4.3 million amino acids.

A breakdown of the main classes is given in the following table:

                   (sub)classes    proteins    AA (x 1000)
DNA transposons        
hAT                      14          361          244
Tc1-Mariner              16          337          136
MuDR                      1          258          150
Maverick / Polinton       1          243          104
En/Spm                    1          110           95
Other                    21          475          280
            
LINEs            
L1                        2          399          337
L2                        2          248          177
CR1                       1          180          130
Other (incl. Penelope)   20          461          381
            
LTR elements        
Gypsy                     1          994          968
ERV                       7          638          444
Copia                     1          401          513
Other                     6          235          245
              
Rolling circles           1           66           85

About 870 of these proteins are entries extracted from the nr GenBank databases. Most of the remaining peptides are translations of interspersed repeat consensus sequences. Many of the coding elements in repeat libraries contain frame shifts and stop codons reflecting errors in the consensus. In order to represent these elements currently some ( 1500 entries ) interrupted coding regions have been translated by using TFASTY or GeneWise, guided by the closest related proteins in the database.

Very similar proteins are represented by only one sequence (e.g. there is only one HIV-1 pol entry). A cut-off of 90% similarity and/or 85% identity has been used, but if many differences appear to be due to ambiguities or errors, more distant matching proteins have been excluded.

The protein database is still very much under development but functions fine for the purpose of masking repetitive DNA that could give rise to spurious matches in translated database searches. As most proteins are classified to the level RepeatMasker classifies transposable elements, it is also already useful for classifying transposable elements. This will be its central role in the interspersed repeat identification program that we are developing. In the future phylogenetic comparisons of transposable elements based on their coding regions can be performed using this database, though we will need to (and will) curate the database significantly before this can be done properly.

Tandem repeats and low complexity DNA

Prior to being compared with WU_BLASTX to the transposable element protein database, the query is checked for the presence of tandem repeats using Tandem Repeat Finder followed by the simple repeat / low complexity algorithm of RepeatMasker. This will avoid false annotations in the comparison against the transposable element proteins. The button allows you to turn off the masking/annotating of low complexity and simple repeats in the final output. Low complexity and simple repeat analysis will still occur prior to looking for matches to the RepeatPep database.

False positives

We have run RepeatProteinMasker on several megabases of genomic DNA and compared it to RepeatMasker output to identify false positive matches. This number could be reduced to zero by eliminating certain low-complexity proteins from the database and reporting matches based on a complexity adjusted alignment score, very similar to that implemented in RepeatMasker. False matches to inverse, non-complimented genomic DNA are also absent, except for an occasional match in very (>70%) GC-rich DNA (no such cases were seen with a score above 30).

A surprisingly large number of genes are derived from transposable element proteins. In 2001 I had identified 50 in the human genome and it is more than likely that other genomes are very similar if not more so, considering that most other genomes are exposed to a much larger variety of transposable elements and DNA transposons in general. So, be aware that some genes may be (partially) masked.


Institute for Systems Biology
This server is made possible by funding from the National Human Genome Research Institute (NHGRI grant # RO1 HG002939-01) 2003.