AMPer: A Database and an
Automated Discovery Tool for Antimicrobial Peptides.
How to use this site.
2007/01/15
Chris Fjell cfjell@interchange.ubc.ca
This site lists the results of a novel method for identifying
antimicrobial peptides (AMPs) from proproteins in the Swiss-Prot
database. We took available, curated AMP sequences and created clusters
of both mature peptide sequence and propeptide sequences, then for each
cluster created a Hidden Markov model. Additional AMPs were recognized
by scanning Swiss-Prot with the HMMs and then looking for agreement
with annotations and similarity with a known AMP by BLAST. Details
will be published soon...
Only the AMPs used in the
clusters are reported here - we'll add the rest as part of a more
comprehensive site as time permits.
Note that the links included in the examples are active.
You can use this site is a couple of ways:
1. You want to look up a particular AMP of interest and see similar
AMPs and information about them.
You can look for the description of the type of AMP in the list of
clusters. (Click a link from the line on the main page "View
cluster from
mature peptide or
propeptide sequences").
Or load the entire set of AMPs (either mature or propeptide) from the
line on the main page "View all mature
peptides or propeptides."
Use the "Find" search tool from your browser to search for peptide
description or sequence. We plan to add a proper SQL-driven search
widget to allow complex queries of the database, but haven't got to
that yet.
2. You have a protein sequence and you want to see if it is similar
to known AMPs using on-line tools.
Select the link from the line on the main page "Enter sequence to
scan for AMPs here".
You will be presented with a box to put the sequence into, and an
E-value cutoff. The E-value cutoff will let you view only the most
significant hits. Enter something small like 1e-10 to just see only
very good hits, or something closer to 1 or higher to see more lower
quality hits.
Be patient. The scan may take several minutes depending on the sequence
and the load on our server. (Remember: you can always download our data
and models and do this yourself.)
You will be presented with four tables. The first two are the HMM
matches to the sequence for the mature clusters and propeptide
clusters. The last two are BLAST comparisons to mature peptides and
propeptides. Two tables will probably be empty since it's unlikely a
sequence will be similar to mature peptides as well as propeptides.
Clicking on a score from a HMM results table takes you to the HMMER
output file for the search.
Clicking on a matching ID from the BLAST
results table takes you to the peptide in the cluster.
Clicking on an alignment length from the BLAST results table takes you to the BLAST results.
3. You have a collection of protein sequences and want to scan against all models using HMMER software.
Get the HMMER software from http://hmmer.janelia.org/.
Put your sequence in a file named, for example, sequence.fa. (I use FASTA format, but HMMER may support others.)
Get the AMPer HMM models from this site (hmms.tgz), download and uncompress them (WinZip on Windows should work; on Linux/Unix/MacOSX use 'tar xzf hmms.tgz'). The models (files ending 'hmm') will be found in subdirectories MATURE and PRO for mature and prosequences, respectively.
Change directory into MATURE or PRO directory and execute the HMMER utility hmmsearch for each model.
For example, there are 146 models for the mature sequences, so on a BASH shell (Unix/Linux/MacOSX) the following will scan the sequence file for each model and put the HMMER output into files named sequence.fa.out1 to sequence.fa.out146:
for ((i=1;i<=146;i++)); do hmmsearch cluster$i.hmm sequence.fa > sequence.fa.out$i; done
To do the same for the prosequence models, change the upper limit since there are 40 models, so the following loop will scan using all models:
for ((i=1;i<=40;i++)); do hmmsearch cluster$i.hmm sequence.fa > sequence.fa.out$i; done
Please consult the HMMER documentation to interpret the output files.
Looking for something you can't find?
E-mail Chris Fjell cfjell@interchange.ubc.ca