inmembrane is a pipeline for proteome annotation to predict if a protein is exposed on the surface of a bacteria.
Installation and Configuration
Dowload the latest version of inmembrane from the github repository: https://github.com/boscoh/inmembrane/zipball/master.
The editable parameters of inmembrane are found in
inmembrane.config, which is always located in the same directory as the main script. If no such file exists, a default
inmembrane.config will be generated. The parameters are:
- the path location of the binaries for SignalP, LipoP, TMHMM, HMMSEARCH, and MEMSAT. This can be the full path, or just the binary name if it is on the system path environment. Use
- 'protocol' to indicate which analysis you want to use. Currently, we support:
gram_posthe analysis of surface-exposed proteins of Gram+ bacteria;
gram_negannotation of subcellular localization and inner membrane topology classification for Gram- bacteria
- 'hmm_profiles_dir': the location of the HMMER profiles for any HMM peptide sequence motifs
- for HMMER, you can set the cutoffs for significance, the E-value 'hmm_evalue_max', and the score 'hmm_score_min'
- the shortest length of a loop that sticks out of the peptidoglycan layer of a Gram+ bacteria. The SurfG+ determined this to be 50 amino acids for terminal loops, and twice that for internal loops, 100
- 'helix_programs' you can choose which of the transmembrane-helix prediction programs you want to use
We provide a number of unit tests for inmembrane:
As inmembrane has a lot of dependencies, these tests are really useful in working out if the dependencies are installed in a way that is compatible with inmembrane. Since not all the binaries are needed, not all tests (and corresponding dependencies) are required for inmembrane to work.
inmembrane was written in Python 2.7. It takes a FASTA input file and runs a number of external bioinformatic programs on the sequences. It then collects the output to make the final analysis, which is printed out and stored in a CSV file.
inmembrane can be run in two modes. It can be run as a command-line program:
python inmembrane.py your_fasta_file
If run in this mode, the CSV will be given the same basename as the FASTA file.
The other way of running imembrane.py is with a custom script, such as
run_example.py where all pertinent input is in the script itself. You can either run this on the command-line like this:
or simply double-click
run_example.py in a file-manager. You can change this by simply duplicating
run_example.py, and editing the parameters in a text editor. In particular, the fields in the parameters include:
- 'fasta' the input FASTA file
- 'out_dir' the directoyr that stores intermediate output
- 'csv' the output CSV file
The output of inmembrane
gram_pos protocol consists of four columns of output. This is printed to stdout and written as a CSV file, which can be opened in spreadsheet software such as EXCEL. The standard text output can be parsed using space delimiters (empty fields in the third column are indicated with a "."). Logging information are prefaced by a '#' character, and is sent to stderr.
Here's an example:
SPy_0008 CYTOPLASM(non-PSE) . SPy_0008 from AE004092 SPy_0009 CYTOPLASM(non-PSE) . SPy_0009 from AE004092 SPy_0010 PSE-Membrane tmhmm(1) SPy_0010 from AE004092 SPy_0012 PSE-Cellwall hmm(GW2|GW3|GW1);signalp SPy_0012 from AE004092 SPy_0013 PSE-Membrane tmhmm(1) SPy_0013 from AE004092 SPy_0015 PSE-Membrane tmhmm(2) SPy_0015 from AE004092 SPy_0016 MEMBRANE(non-PSE) tmhmm(12) SPy_0016 from AE004092 SPy_0019 SECRETED signalp SPy_0019 from AE004092
the first column is the SeqID which is the first token in the identifier line of the sequence in the FASTA file
the second column is the prediction, it is CYTOPLASM(non-PSE), MEMBRANE(non-PSE), PSE-Cellwall, PSE-Membrane, PSE-Lipoprotein or SECRETED. Any 'PSE' (Potentially Surface Exposed) annotation means that based on the predicted topology, the protein is likely to be surface exposed and will be protease accessible in a membrane-shaving experiment.
the third line is a summary of features detected by external tools:
- tmhmm(2) means 2 transmembrane helices were found by TMHMM
- hmm(GW2|GW3|GW1) means that the GW1, GW2 and GW3 motifs were found by HMMER hmmsearch
- signalp means a secretion signal was found SignalP
- lipop means a Sp II secretion signal found by LipoP with an appropriate CYS residue at the cleavage site, which will be attached to a phospholipid in the membrane
the rest of the line gives the full identifier of the sequence in the FASTA file.
As it is the nature of bioinformatic programs that they are changed and updated severely with different versions, stable APIs with consistent output formats are the exception rather than the norm. It is very important that you have the exact version that we have programmed against.
Required dependencies, and their versions:
- TMHMM 2.0 or MEMSAT3
- SignalP 4.0
- LipoP 1.0
- HMMER 3.0
These instructions have been tailored for Debian-based systems, in particular Ubuntu 11.10. Each of these dependencies are licensed free to academic users.
Only one of TMHMM or MEMSAT3 are required, but users that want to compare transmembrane segment predictions can install both.
- Download and install TMHMM 2.0 from http://www.cbs.dtu.dk/cgi-bin/nph-sw_request?tmhmm.
- Download SignalP 4.0 http://www.cbs.dtu.dk/cgi-bin/nph-sw_request?signalp. You will need to fill out the form with an institutional email address and accept the academic license. The software will be emailed to you.
- Follow the installation instructions at http://www.cbs.dtu.dk/services/doc/signalp-4.0.readme.
- Download HMMER 3.0 from http://hmmer.janelia.org/software.
- The HMMER user guide describes how to install it. For the pre-compiled packages, this is as simple as putting the binaries on your PATH.
- Download LipoP 1.0 from http://www.cbs.dtu.dk/cgi-bin/nph-sw_request?lipop. The installation proceedure is similar to that for SignalP.
- Download MEMSAT3 from http://bioinfadmin.cs.ucl.ac.uk/downloads/memsat/memsat3/ (only memsat3_academic.tar is required).
MEMSAT3 requires NCBI BLAST ("legacy" BLAST, not BLAST+) using the SwissProt (swissprot) database.
- Legacy BLAST can be downloaded at ftp://ftp.ncbi.nlm.nih.gov/blast/executables/release/LATEST/ installed using the instructions provided by NCBI http://www.ncbi.nlm.nih.gov/staff/tao/URLAPI/unix_setup.html. We have tested with version 2.2.25.
- You will need both the 'nr' database and the 'swissprot' database, since 'swissprot' is indexed against 'nr'. (The other option is to download the FASTA version of Uniprot/Swiss-Prot from ftp://ftp.uniprot.org/pub/databases/uniprot/current_release/knowledgebase/complete/uniprot_sprot.fasta.gz and create your own BLAST formatted database with using the BLAST formatdb tool).
Edit the runmemsat script included with MEMSAT3 to point to the correct locations using absolute paths:
- 'dbname' is the location of your BLAST formatted swissprot database
- 'ncbidir' is the base directory of your BLAST installation
- 'execdir' is the path where the MEMSAT3 executable resides
- 'datadir' is the the path to the MEMSAT3 data directory )
Note the the 'runmemsat' script refers to PSIPRED v2, but it means MEMSAT3 - PSIPRED is not required.
(*) We provide a few third party Python libraries ( Beautiful Soup, mechanize and twill) and Suds ) in the inmembrane distribution so there is one less hoop to jump through for installation. If you are an expert user who would, for some reason, like to use your own system copies of these libraries (eg those installed via apt-get or pip install), you can simply rename or delete the approriate library directories in the inmembrane distribution.
It is a fact of life for bioinformatics that new versions of basic tools changes output formats and API. We believe that it is an essential skill to rewrite parsers to handle the subtle but significant changes in different versions. We have written inmembrane to be easily modifiable and extensible. Protocols which embody a particular high level workflow are found in
All interaction with a specific external program or web-site have been wrapped into a single python plugin module, and placed in the
plugins directory. This contains the code to both run the program and to parse the output. We have tried to make the parsing code as concise as possible. Specifically, by using the native Python dictionary, which allows an enormous amout of flexibility, we can extract the analysis with very little code.
inmembrane development style guide:
Here are some guidelines for understanding and extending the code.
- Confidence: Plugins that wrap an external program should have at least one high level test which is executed by run_tests.py. This allows new users to immediately determine if their dependencies are operating as expected.
- Interface: A plugin that wraps an external program must receive a params data structure (derived from
inmembrane.config) and a proteins data structure (which is a dictionary keyed by sequence id). Plugins should return a 'proteins' object.
- Flexibility: Plugins should have a 'force' boolean argument that will force the analysis to re-run and overwrite output files.
- Efficiency: All plugins should write an output file which is read upon invocation to avoid the analysis being re-run.
- Documentation: A plugin must have a Python docstring describing what it does, what parameters it requires in the
paramsdictionary and what it adds to the
proteinsdata structure. See the code for examples.
- Anal: Unique sequence ID strings (eg
gi|1234567) are called 'seqid'. 'name' is ambiguous. 'prot_id' is reasonable, however conceptually a 'protein' is not the same thing as a string that represents it's 'sequence' - hence the preference for 'seqid'.
- Anal: All file handles should be closed when they are no longer needed.