Reference APSES domains (reference species)

Multi FASTA file of all APSES domains in fungal proteins.

Generating this file turned out to be much more difficult than anticipated:

One feature I want this file to have is headers that identify the yeast protein that each ASPES domain is most similar to. E.g. if the ASPES domain of Aspergillus terreus protein ATEG_06370 is most similar to that of yeast Xbp1, I want it to be called Xbp1_ASPTE. Such meaningful protein names - rather than abstract GI or RefSeq identifiers - will be extremely helpful when it comes to analysing conservation or phylogeny.

The other feature is of course obtaining the full-length sequence of the ASPES domain. Unfortunately BLAST being a local-alignment algorithm does not guarantee that it will not shave off significant bits and pieces from the termini.

At this point, I would normally start writing a program. But let's see how far we can get with tools we find on the Web.

Defining the most similar ASPES domain in yeast

Normally we would find the most similar protein in another species by executing a BLAST search. In our case however, we have 70 sequences. Doing this by hand is possible - but painful. Even clicking through the precomputed BLinks (that we would conveniently find on the page returned through "Get selected sequences") will not help us, since, we are not looking for the most similar protein per se, but for the most similar ASPES domain. So what we need is (1) an input file of ASPES domain sequences, and then (2) a way to BLAST them against the yeast genome. Let's ignore for the time being the requirement for full-length domain sequences and stick with the HSPs that PSI-BLAST has found. Parsing the BLAST file and extracting the sequences by hand is, again, possible, but painful. Fortunately there is a simpler way. If we navigate to the page from where we sent the BLAST query, we have several options to display search results:

• Pairwise: the default
• Pairwise with identities: showing only differences to the query sequence
• query anchored with/without identities: looks something like a multiple sequence alignment, hyphens for gaps, insertions relative to the query are displayed below the sequence
• flat-query anchored with/without identitites: This now looks like a multiple sequence alignment (in fact it is one - all sequences aligned to the profile).
• hit-table: this gives only the numerical parameters describing the quality of the matches.

Thus using the flat-query anchored with/without identitites option, it is reasonably straightforward to obtain the matched sequences, copy and paste them into a Word document and convert that into a multi-FASTA format with a few Edit > Replace commands.

The ASPES domain sequences

This requires to take all sequence identifiers, use their APSES domains and search them against the yeast genome. I actually have given up on finding a Web-tool to do this. Of course this can be done manually, through Blinks - but having to do this for 70 sequences was an uninspiring prospect. And it seems there are no BLAST Webservices that will accept batch-input of lists of sequences.

The full-length protein sequences were copied from the previously prepared input file of [[All_APSES_proteins|86 proteins] and pasted into the input form of the EBI ClustalW service. While this is no longer considered state-of-the-art for multiple sequence alignments, it is computationally efficient and sufficiently accurate for the purpose of approximate domain boundary definition. What we want to construct an input file for aligning just the APSES domains: this should contain the following

• our yeast APSES domain (this defines the boundaries of the domain we are interested in)
• enough sequence extending it N- and C-terminally for the other proteins to ensure we are not throwing out conserved amino acids
• but not too much, since irrelevant sequence can cause problems for the alignment.

Scrolling through the ClustalW result page, the alignment blocks containing the Mbp1 APSES domain sequence were copied and pasted into a MSWord test document, then manually edited to contain only the APSES domains plus some 10 or 20 residues on each end. Through some simple replace commands, this was then brought into a FASTA format. What's a bit annoying is that this changes the headers to contain only the first word (in our case mostly the GI number) .. i.e. from a FASTA input of ...

>6320147 NP_010227.1 Mbp1p [Saccharomyces cerevisiae]
MSNQIYSARYSGVDVYEFIHST...

... we get a Clustal record of ...

6320147         --------------------------------------MSNQIYSARYSGVDVYEFIHST 22

...which we can change back into a FASTA record:

>6320147
MSNQIYSARYSGVDVYEFIHST

Tuhs losing part of the header information. There is no easy way to repair the headers in MSWord, but using a trivial perl program this can be automated:

However I consider this cosmetics - the file would have been just as valid with only the GI numbers in the header. Here is the resulting FASTA file containing only APSES domains:

Sources

• [1]
• [2]
• [3]