Hi Joni. I'm seconding Geoff's email - it's really good to hear you're progressing with adding services to JABAWS ! I've been working on adding JABAWS2 services to Jalview for the v2.8 release - and I think we might be able to help each other out here...
On 15/11/2011 19:48, Joni Fazo wrote: > I'm working to create a new jabaws web service for the tool gblocks > (http://molevol.cmima.csic.es/castresana/Gblocks.html). I am using > the MsaWS interface. I have the web service working - sort of :) > > My problem is that gblocks expects aligned sequence as input. My > gblocks web service works fine when given input sequence that is all > the same length, with no gaps. But otherwise it fails, reporting that > the sequences are of different lengths. All gaps seem to be stripped > off the sequence prior to the web service being called. > > I see that the Jalview class MsaWSClient has an attribute "submitGaps" > with a default value of false. Is there anyway to set this to true > when calling a specific web service? > > This class also has a method canSubmitGaps with a TODO that makes me > think this capability is not yet available. (see below). Any tips on > a work-around? > > private boolean canSubmitGaps() > { > // TODO: query service or extract service handle props to check if > we can > // realign > return (WebServiceName.indexOf("lustal") > -1); // cheat! > } > There's no straightforward workaround for version 2.7 of Jalview, I'm afraid. If you take a close look at the execution steps leading to a job being submitted by Jalview, you'll see that the 'canSubmitGaps' method is not actually used to decide if gaps are removed, but instead used to decide if a 'Realign with ..' menu entry is generated in addition to the standard 'align with ..' menu entries for alignment web services. However, hopefully, you've developed your service with the JABAWS 2 development code that Peter Troshin sent you, in which case there's a definite way forward... The real problem here is that GBlocks isn't a multiple alignment service, and trying to use code that expects to be talking to one kind of program but instead does something else is simply not going to work. The GBlocks service is a specific variant of an alignment analysis service where you submit an alignment and get back annotation of some kind; which, in this case, is annotation indicating which columns should be hidden in the view. Implementing a service for GBlocks following this pattern will ensure that it is robust and ultimately more usable, since it does what it says on the tin. Fortuitously, JABAWS 2 already includes a canonical alignment annotation service model that would suit GBlocks, and I have already coded a client in the Jalview development branch[1] that could be easily adapted to mark the regions on the alignment annotated by the service as hidden regions. All that would be needed is for you to refactor your new service code to implement the compbio.data.msa.SequenceAnnotation interface, and then add some semantic sugar in order to inform Jalview that the annotation should be used to mark the given regions as hidden. The semantic sugar comes in two forms: 1. Label the service with an appropriate service category in the JABAWS2 discovery service. GBlocks performs a specific form of conservation analysis to identify conserved regions in the alignment, so it could be put in the 'Conservation' category - but this contains methods that produce alignment quality and conservation histograms. Instead, I'd suggest creating another category, such as 'Conserved Region Detection'. This will mean it gets placed in its own submenu on the web services menu. 2. Label the results in a way that makes it clear what Jalview should do with them. The result from an annotation service is a series of named sets associated with each sequence or the alignment as a whole, and each set consists of either a series of scores for each column of the reference (sequence or alignment) and a series of intervals which may optionally have scores associated with them. In the case of a single Gblocks run, you'd return the conserved regions as intervals in a single score set associated with the whole alignment. I'd recommend calling that score set something obvious like 'CONSERVED_REGION' .. that way it's quite clear what the interval refers to, and isn't specific to GBlocks - in case other conserved region detection programs get added to JABAWS. The point of the sugar is to allow Jalview to do something sensible with the results. Typically, phylogenetically informative columns get marked with a '*' - so Jalview could add an annotation row where a '*' is present within each column of the conserved region. It could also automatically hide the columns outside the regions in the alignment view that GBlocks identified. I'm happy to code the client that would do this, if you'd be happy to refactor the Gblocks service ! Jim. [1] The jalview development branch build is at http://www.compbio.dundee.ac.uk/users/ws-dev1/jalview/develop/, and you can see the branch history at http://source.jalview.org/gitweb/?p=jalview.git;a=shortlog;h=refs/heads/develop _______________________________________________ Jalview-dev mailing list [email protected] http://www.compbio.dundee.ac.uk/mailman/listinfo/jalview-dev
