What Is A Good Web Front End For (Blast) Homology Search?

Question

[Ketil]

One fairly common task is to make some sequences available for BLAST search over the web. So far, this has meant installing the NCBI web front end to BLAST, but this is something like 15 years old, and it shows. Although we have some local enhancements that allows the user to click on sequence names to download them, etc, I really would like a more flexible replacement. Surely something more advanced exists? Something equally easy to download and set up?

Edit: To be more specific, I would like a user friendly (i.e. simple) interface to do sequence search (not necessarily BLAST) against our own sequence data. Features I'd like to see might be

1. visualization of the alignments (better than the rudimentary NCBI imagemap)
2. links to target sequences (or to a genome browser or similar)
3. hit selection/filtering based on e.g. taxonomy
4. free software with a sensible (GPL, BSD, MIT..) license
5. simultaneous search against multiple databases
6. automatic database discovery (no more perl config_setup.pl)
7. automatic restriction of available alignment algorithms based on sequence types (no more blastp on nucleotides sequences)

(Feel free to suggest additions to the list in the comments below!)

[Yannick Wurm]

Hey Ketil, please check out Anurag's answer below (https://github.com/yannickwurm/sequenceserver). If fullfills 2,4,5,6,7. Advanced blast params should make 3 possible (but would require smart users). www.sequenceserver.com remains to be updated as of 2011.06.06)

Answer 1

[Anurag Priyam]

You can have a look at SequenceServer too. Its easy to deploy, and has an intuitive UI. Currently in 'Beta 1' it supports:

• simultaneous search against multiple databases
• automatic database discovery (no more perl config_setup.pl)
• automatic restriction of available alignment algorithms based on sequence types (no more blastp on nucleotides sequences)
• links to target sequences (or to a genome browser or similar) - linking to a genome browser is currently being worked on.

On TODO is:

• visualization of the alignments (better than the rudimentary NCBI imagemap)

And SequenceServer is open source.

[Anurag Priyam]

Oh, and I forgot to mention you can enter other (advanced) parameters too as you would from the command line. So you can blast with different evalue settings, or change num_alignments.

[Yannick Wurm]

(FYI there's been quite a few updates since this...)

Answer 2

[Pierre Lindenbaum]

If by 'user' you mean some skilled bioinformaticians, I would create a WEB-Service (WSDL+SOAP/REST) to let them play with your data as they want without having to connect to your HTML interface. You can find this kind of service thought the biocatalogue.org . For example: http://www.biocatalogue.org/services/47

If you really need a new kind of web interface (why?), I would play with a nice javascript framework (GWT, ...) to create an interactive experience.

[Ketil]

No, the users are primarily biologists, and they need a clickable interface, not a programmable one.

Answer 3

[User 59]

If you're presenting in-house tools with a web interface you might want to look at Mobyle, the successor to PISE.

This provides not only an interface to BLAST but also a number of other bioinformatics tools. Can't vouch for its ease of installation however. You can download it here

[Ketil]

Mobyle looks interesting, but the blast service doesn't seem to be very advanced - it merely dumps blast's text output. Am I missing something?

[User 59]

No, I guess not, it's there to provide an interface to BLAST, I replied to this before you updated your answer. Your feature list doesn't seem unreeasonable, but you're really talking about post-BLAST processing. I think the point about Mobyle is that you could, if you so desired, customise it for your requirements.

Answer 4

[Ketil]

After looking at Mobyle, I think perhaps also Galaxy might be worth a look. These systems are comprehensive and general, which is an advantage if you plan to grow and customize, but maybe not the small and simple solution.

Answer 5

[Pascal]

ViroBlast is a easy to set up web interface for BLAST. Compared to NCBI's web front end it is only slightly more flexible, but at least it looks more modern. It can be obtained for free for academic use.

http://bioinformatics.oxfordjournals.org/content/23/17/2334.abstract

[Ketil]

This is fairly close, although not terribly feature laden. It seems to cover #7, and provides links to genbank - it's unclear if I can link to my own sequences, and there's no visualization. I'd prefer something with a standard license, but I guess it's not a show-stopper.

[Ketil]

It still sucks that you need to manually specify the databases for blastn, tblastn, etc etc - even after putting nucleotide and protein databases in separate directories. However, I'm calling this the answer, since it seems to be the option that sucks the least.

Answer 6

[Neilfws]

You might like to look at Pise. It allows you to create web interfaces to any command-line program including BLAST, quickly, using config files in XML format. It comes with scripts that automate the build, given the command-line arguments.

It's been around for a while so I don't know whether it works with the very latest BLAST. The interface is quite basic - not a whole lot more than you'd get from the NCBI web package, but may have more options for customization.

Another option - depending on how much work/time you want to invest - is to roll your own, which is very simple these days with modern web frameworks such as Rails (Ruby) or DJango (Python).

I'd also suggest that you consider whether provision of web interfaces is a good use of your time. It's quite a lot of work and maintenance (of the boring sysadmin variety), just to make life a little simpler for a few users who probably won't even appreciate it. Sometimes I think it would be better (as in more useful and educational) to set up BLAST on a server, provide SSH access and run a short course/tutorial on the basics: logging in, running a command-line BLAST and saving the results, then provide some basic parsing scripts.

[Ketil]

There's a lot of (primarily) biologists who just want to look for their favorite gene in some data. The most likely alternative is that they send me their sequences in an email, and that I have to run blast for them... which is not a good use of my time :-)

[User 59]

I think you mean 'send me their sequences in a Word document attachment in an email' :)

[User 59]

I see we had the same thought ;) But PISE is superseded by Mobyle now https://projets.pasteur.fr/wiki/2/Download

[Neilfws]

Good to know. Years since I used Pise.

[Michael]

Sometimes they might do it in a better and more standardized way and use ... an Excel sheet

[Ketil]

Suffice it to say that any solution can not be overly technical and complicated...