TMbase - A Database of Membrane Spanning Protein Segments

TMbase - what is it?

TMbase is a database of transmembrane proteins and their helical membrane- spanning domains. TMbase was originally meant as a tool for analyzing the properties of transmembrane proteins. It is intended to facilitate the following tasks:

finding positional preferences of certain amino acids
deriving an improved method for the prediction of transmembrane domains
testing of such prediction schemes
statistical testing of general hypotheses concerning transmembrane proteins

TMbase is mainly based on SwissProt, but contains informations from other sources as well. All data is stored in different tables, suited for use with any relational database management system. These tables are distributed as ASCII files. Since the creation and internal maintenace of TMbase is done on a PC using MS-ACCESS, the ACCESS version of TMbase is available as well.

The current release of TMbase (TMbase25) is based on SwissProt release 25. It has been briefly described in the meeting abstract:

K. Hofmann, W. Stoffel
TMBASE - A database of membrane spanning protein segments
Biol. Chem. Hoppe-Seyler 374,166 (1993)

The data stored in TMbase contain, among other things, information on the following:

What SwissProt entries are transmembrane proteins?
How many transmembrane domains do they possess?
Where in the sequence are the TM-domains located?
What is the sequence of the (putative) membrane-buried parts?
What is the sequence of the flanking regions?
What is the putative orientation of the transmembrane proteins?
What is the putative orientation of the TM-helices?
To what type of membrane are the proteins associated?
To what species/taxonomic group do the proteins belong?
What transmembrane proteins are related to another?
What is the 'relative degeneracy' of a TMbase entry?

All these data are present in a form that can be easily queried. For a more detailed description of the file structures, see the paragraphs below.

Most of the SwissProt-based information has been extracted from theSwissProt annotation in a semi-automatical manner. If there are any errors in these annotations, they will probably be propagated in TMbase, too. However, some consistency checks have been performed though, and in the case of any inconsistencies the contradictory information has not been included into TMbase.
It should be noted, that errors in the annotation of transmembrane helices are quite frequent since this reflect only the uncertainty of TM-domain prediction and the lacking of experimental data.

The methodology for grouping the transmembrane proteins into families and for calculating the 'relative degeneracy' is explained in the paragraphs below.

Explanation of the individual files:

All files are given as comma-delimited ASCII files. Fields that contain character strings are enclosed in double quotes.
The first record in each file holds the field names instead of values. For import from some database systems, this first record has to be deleted first. Some database systems, however, can use this record for automatically assigning field names.

Besides the individual TMbase tables in the *.txt files, the whole TMbase collection is present as a compressed tar-file (tmbase25.tar.Z) for use on unix-machines or as a compressed zip-file (tmbase25.zip) for use on PCs.

tmb_e: (TMbase entries)

This file has one record for each transmembrane protein contained in SwissProt25.
The records have the following structure:

counter     : integer  : just an index
ID          : string10 : the SwissProt ID field
Description : string80 : the SwissProt DE field (possibly truncated)
Length      : integer  : the length of the protein
No of TM    : integer  : the number of annotated TM-helices

tmb_h (TMbase helices)

This file contains one record for each transmembrane helix that is annotated in SwissProt25.
The records have the following structure:

helix-ID    : string15 : unique Identifier for a helix
ID          : string10 : the associated ID of the tmb_e entry
TM#         : integer  : position-number of the helix in the protein
Start       : integer  : position of first residue in the helix
Stop        : integer  : position of last residue in the helix
Pre         : string5  : the five residues N-terminal of the helix
Transmem    : string35 : the amino acid sequence of the helix
Post        : string5  : the five residues C-terminal of the helix
Comment     : string41 : any comments (like e.g. 'signal')

tmb_h2 (predicted TMbase helices)

This file contains one record for each transmembrane helix that is predicted by TMpredict.
The records structure is identical to that of tmb_h with one additional field at the end:

Orientation : string2  : io=inside to outside, oi=the opposite
                         inside'  means normally the cytoplasmic face
                         outside' the lumenal face of the membrane
                         depending on the organelle (see tmb_mem)

tmb_or (transmembrane protein orientation)

This file contains one record for each transmembrane protein, giving the annotated orientation of the protein. The orientation has been automatically deduced from either annotation of the orientation itself or from annotation of features like glycosylation sites or phosphorylation sites. Only entries with consistent annotations have an orientation assigned.
The records have the following structure:

ID          : string10 : the SwissProt ID field
Nterm       : string1  : either 'i' for inside or 'o' for outside
                         position of the N-terminus of the protein.

tmb_hor (transmembrane helix orientation)

This file contains one record for each transmembrane helix, giving the annotated orientation of the membrane spanning segment. The helix orientations have been derived from the protein orientation given in tmb_or using the position number of the helix and the assumption of an alternating pattern of i->o and o->i helices.
The record structure is similar to the one in tmb_or:

helix-ID    : string15 : the helix identifier used in tmb_h
Nterm       : string1  : either 'i' for inside or 'o' for outside
                         position of the N-terminus of the helix

tmb_mem (membrane types the TM-proteins are associated with)

This file has one record for each transmembrane protein, giving a classification of the membrane type and subtype.
The following abbreviations are used for membrane types and subtypes:

Organelle-type:    pro = prokaryotic cell membrane
                   er  = eukaryotic ER-, Golgi- or Plasma-membrane
  	           mit = mitochondrial membrane
   	           chl = chloroplast membrane
		   vir = misc. viral membrane proteins

Subtype       :    er  = ER membrane
	           gol = Golgi membrane
		   mic = unspecified microsomal proteins
		   pla = euk. plasma membrane
		   lys = lysosomal membrane
		   inn = inner membrane (e.g. mitochondrial)
		   out = outer membrane  
		   thy = thylakoid membrane
		   env = envelope membrane (viral)

The records have the following structure:

ID             : string10 : the SwissProt ID field
Organelle-type : string3  : the organelle the protein is associated to
Subtype        : string3  : the subtype of the membrane (see above)

tmb_fam (grouping of proteins into families/superfamilies)

This file contains one record for each transmembrane protein, indicating its membership to a similarity family. Based on considerations described below, two different sets of families have been created. One set is called 'PAM80-families', the other is called 'PAM200-families'. The families of the latter set tend to be bigger because they group more distantly related proteins into one family. The 'PAM200-reduced' set of families is very similar to the 'PAM200-families', but some obviously wrong groupings have been removed by hand.
The records have the following structure:

ID            : string10 : the SwissProt ID field
PAM80-Family  : integer  : the number of the family the protein belongs to
PAM200-Family : integer  : ditto
PAM200-reduced: integer  : ditto

tmb_200r ('PAM200-reduced' family statisctics and descriptions)

This file contains a record for each family of the 'PAM200-reduced' set. For each family the number of members and a short description are given. At the time being, descriptions are only supplied for families with 5 or more members. If a consensus transmembrane topology of a familly could be determined, the probable number of TM-helices is indicated.
The records have the following structure:

PAM200r-Family : integer  : the family number (see tmb_fam)
population     : integer  : the number of members in the family
function       : string80 : short description of the family
No of TM       : string10 : No. of TM-helices of a typical member

tmb_deg (degeneracy values for transmembrane proteins)

When extracting properties from a set of proteins, it is important to account for the occurence of similar proteins, biasing the statistics towards frequently sequenced proteins. This file contains several measures for 'degeneracy' for each transmembrane protein. For a description of the degeneracy values, see below.
The records have the following structure:

ID                    : string10 : the SwissProt ID field
all                   : real     : Degeneracy measures, type II
known orientation     : real     : 
ER, known orientation : real     : 
all/PAM80             : integer  : Degeneracy measures, type I
all/PAM200            : integer  :
ER/PAM200             : integer  :

tmb_spec (species (organism) of the TMbase entry)

This file contains one entry for each transmembrane protein, splitting the ID into two sub-fields. One subfield characterizes the protein, the second one the organism. As a basic taxonomic information on the species, the one-letter code (P=prokaryotic, E=eukaryotic, V=viral) and the taxonomy code and number of the SwissProt file SPECLIST.TXT are included.
The records have the following structure:

ID        : string10 : the SwissProt ID field
Protein-ID: string4  : The part of the ID representing the protein
Species-ID: string5  : The part of the ID representing the organism
Group-ID  : string5  : taxonomy ID in the SwissProt SPECLIST.TXT
Spec-Type : string1  : one letter code for basic taxonomy
Spec-No   : integer  : taxonomy number in the SwissProt SPECLIST.TXT

tmb_taxo (taxonomic classification in details)

In order to use a more detailed taxonomical classification, this file contains for each Group-ID (in tmb_spec) the complete taxonomy given in the bottom part of the SwissProt file SPECLIST.TXT
The records have the following structure:

Group-ID  : string5  : taxonomy ID in the SwissProt SPECLIST.TXT
Spec-Type : string1  : one letter code for basic taxonomy
Spec-No   : integer  : taxonomy number in the SwissProt SPECLIST.TXT
1st level : string27 : the levels of taxonomy
2nd level : string33 :        .
3rd level : string41 :        .
4th level : string37 :        .
5th level : string22 :        .
6th level : string16 :        .
7th level : string19 :        .

tmb_ncs (number of chains and signal sequences)

If the information contained in TMbase is used in evaluation TM-protein predictions, is might me necessary to have data too: SwissProt entries contain in may cases (annotated) N-terminal signal sequences that are not present in the mature protein. These sequences share several feature with transmembrane segments and could possibly confuse prediction programs.
Several SwissProt entries contain polyproteins that are cleaved into individual chains posttranslationally. If this happens to transmembrane proteins, the assumption that an inside-to-outside TM-segment is followed by an outside-to-inside segment is not necessarily true.
This file contains for each of the TMbase protein the number of annotated chains (0 if no annotation is present) and the number of annotated signal sequences (either 0 or 1). A chain number of 0 normally can be treated equal to a chain number of 1.
The records have the following structure:

ID           : string10 : the SwissProt ID field
No of chains : integer  : the number of annotated chains
No of signals: integer  : the number of annotated N-terminal signals.

Considerations of TM-protein Families and Degeneracy

Since TMbase was originally meant for the extraction of statistical data, it became important to avoid (or reduce) any bias that may be caused by the existence of several similar sequences in the database. For accomplishing that, several different methods were used, either based on similarity of the membrane spanning domains themselves or on similarity of the whole protein.

The similarity method that is used in the files tmb_fam and tmb_deg is based on the following procedure:
The 'allalldb'-function of the DARWIN server at the ETH-Zurich was used to get similarity data for each possible pair of transmembrane proteins that exceeds a certain similarity threshold. These similarity data include a Smith-Waterman score and an estimated PAM value. (PAM means the number of assumed point mutations per 100 residues).
In the next step, the proteins were grouped into families of sequences that have a PAM distance of less than 80. Each protein that has a PAM value of less than 80 to any member of an existing family was assumed to belong to that family, too. This method does not implies that each pair of sequences in a particular family must necessarily have a PAM distance less than 80. The families resulting from this procedure are called 'PAM80-families'.
In a totally analogues fashion, PAM200 families were computed. These PAM200 families include even more distantly related sequences and, as a consequence, there are fewer of these families, each containing more sequences.

Based on this family groupings, the degeneracy values 'all/PAM80', 'all/PAM200', and 'ER/PAM200' have been calculated. The 'all/PAM80' field for a given protein holds an integer number of family members belonging to the same family as the original sequence. The reciprocal value of this field could be used as a weighting factor in statistical analyses. The 'all/PAM200' field holds an analoguous value assuming PAM200 families. The 'ER/PAM200' field, that is present only for proteins belonging to the organelle type 'er' (see tmb_mem), contains the number of 'ER'-proteins belonging to the same PAM200-family as the original protein.

The similarity measure above includes only information of family-membership, each other member of the same family contributes by a value of '1', no matter if it is a close relative or a more distantly related sequence. To avoid this restriction, the 'typeII' degeneracy measure has been introduced. The 'type II' degeneracy of a given protein i, Deg(i), is given as the following sum

                            1
   Deg(i)=      Sum  --------------
                 j   1+PAMdist(i,j)

where j is each of proteins belonging to the same family. This too is clearly not a perfect measure for protein degeneracy. Nevertheless, it has proved to be useful in avoiding large-family bias in several statistical analyses.
The 'type II degeneracies' has been calculated for three different sets of proteins. One column in tmb_deg is calculated for all proteins, one for proteins with known transmembrane orientation, and a third one for a subset thereof, containing only ER-type proteins.

Availability

At the moment, TMbase25 is available by anonymous ftp from
ftp.isrec.isb-sib.ch in the directory pub/tmbase
or from
ncbi.nlm.nih.gov in the directory repository/TMbase

How to contact the author

If there are any further questions , feel free to contact the author under this address:

Kay Hofmann
Bioinformatics group
ISREC
Chemin des Boveresses 155
CH-1066 Epalinges s/Lausanne
Switzerland
-------------------------------
FAX: +41 (21) 652-6933
Email: khofmann@isrec-sun1.unil.ch