Re: [Gusdev-gusdev] LoadGeneFeaturesFromXML error

[Jonathan C. <cra...@pc...>]

Hi Michael-

MICHAEL LUCHTAN wrote:
> Some more information:
> Error occurs in method setAlgId, which sets name to 'Pf Annotation'.
> So when running this module with --verbose, there is an error like:

This is a relatively straightforward case of a plugin making a (perhaps
un- or not-so-well documented) assumption about the initial state of the
database.  In this case, the plugin expects to find an entry in
core.Algorithm with the name "Pf Annotation".  We use the Algorithm table
to represent not only GUS plugins, but also "external" programs/algorithms
like Genscan, BLAST, etc.  For gene predictions in particular, the
GeneFeature table has a prediction_algorithm_id which we use to indicate
the gene predictor that generated the gene model.  In the PlasmoDB project
we have an Algorithm called "Pf Annotation", which is not a formal algorithm
per se, but rather a placeholder to indicate that the gene model in question
is an "annotated gene" (i.e., it may have come from an automated gene
prediction program initially, but it has since been modified and/or reviewed
by an annotator, and then released as part of the official annotation.)

> Do I need to change line 216 in LoadGeneFeaturesFromXML from
>  my %alg = ( name => 'Pf Annotation' );
> to
>  my %alg = ( name => 'GUS::Common::Plugin::LoadGeneFeaturesFromXML' );
> ??

No; you should instead add an appropriate entry into the Algorithm table.  If
you're loading official Plasmodium falciparum gene predictions, then you should
make an entry for "Pf Annotation"; otherwise you should enter whatever best
describes the source of the genes that you're loading.  I know that we have
used the same plugin to load gene models from Plasmodium yoelii, so it
shouldn't be hard-coded to look for "Pf Annotation", but it's possible that
this oversight hasn't been fixed yet.  If it hasn't been done already, the
name of the algorithm should be made into a command-line parameter to the plugin.
The plugin should also behave better if the requested Algorithm cannot be
found (and it should perhaps even create a new entry, although that could lead
to trouble once the inevitable spelling mistakes are made.)

Jonathan