[Bioc-devel] homolog.db package

Hi Luigi,

I am the one responsible for trying to maintain order with the
annotations in this project.  To date, there has not been a lot of
external interest in cross-species mappings.  Hong built the  
homolog.db
package that you found, but he didn't update these packages for the  
most
recent release, so unless Hong speaks up rather soon, it really seems
that might be more or less abandoned at this point.  I have built the
inparanoid packages in anticipation that someone like you might come
along some day, but it is difficult to predict the future, let alone
write software for it, which is why there are not a lot of functions  
to
make use of these packages yet.  I have experimented with some of  
these
mapping problems, but they are not simple problems as the mappings
frequently tend to be many to one or many to many.  And layered on top
of that is the fact that the inparanoid project uses inconsistent  
labels
for protein sequences which means that you have additional mapping  
steps
to do once you find a match.  Anyhow the way that you want to handle
this mapping will depend almost entirely on the context of what
questions you are asking.  If you have a particular use case in  
mind, or
a specific need, lets discuss it offline and see what can be done  
about
it.  It might be that we can add some methods to improve things.  The
inparanoid packages themselves are due for a major overhaul very  
soon as
the sources have very recently undergone a major revision.

 Marc




Tony Chiang wrote:

Let's keep this onlist so others can also respond. I don't know about
conference calls...I am certainly not apart of any if there are. I
completely agree with you about efficiency and usability. It might be
worthwhile to talk to the maintainers of the packages to see if  
methods
already exist. You could always create your own package that  
depends on the
.db packages and submit to Bioc.

On Fri, Nov 6, 2009 at 12:45 AM, Luigi Marchionni  
<marchion at jhu.edu> wrote:

Is there any conference call or something like that where  
developer talk
and discuss this?
I do not want to annoy anyone, though I think I have a perspective  
about
annotation that can contribute overall.
I am a professional annotator, so to speak, from sequences, to  
ids, to
ontologies, and I am a biologist, and R end user.
My perspective is that something must be computational efficient,  
AND
usable.
I am not saying there is the need to change the packages, but that  
there is
the need to provide methods.
Maybe I am not aware of them if they exist (that's why I ask), but  
I can
definitely can tell you what biologist-not-scared-of-R needs.
One line of code to a bunch of tasks.
like:

ReadAffy()

imagine:

mapEntrez(ids,'Mmu','Hsa')

The latter is what my code does.
If I have to recode everything to make it work with maintained bioc
metadata packages I'll do it.
However, still remains the problem of where such methods should sit.
A new package?
annotate?
AnnotationBbi?

Thanks, and now I go to bed.
Luigi

On Nov 6, 2009, at 3:10 AM, Tony Chiang wrote:

Hi Luigi,

I was not criticizing...I just thought I might point you to some  
other
packages. I believe that the annotation packages do have methods  
that allow
for the translation fairly easily as these are now SQL databases  
(please
correct me if I am wrong anyone). I don't maintain these packages. I
certainly would not mind if someone were to write some simple  
methods or
functions that wrapped around these data packages.

There is always the problem of the many to many mapping  
though....and this
might be why the annotation package is used rather than a flat file.

Cheers,
--Tony

On Fri, Nov 6, 2009 at 1:02 AM, Luigi Marchionni  
<marchion at jhu.edu> wrote:

Thanks.
I am a good citizen overall.
I see now where all I needed is sitting.
Though, metadata packages come with a drawback (I just expecting  
to be
wrong again), they do not contain methods to do stuff.
I really would like to provide people bioC compliant ways to run  
one line
of code and get mapping done.
This can go in any "software" package I am not aware of, however  
this is
going to make the difference.
I do not want to change the way things are, I want to make them  
works
easily to the end user  (a non-scared-by-R biologist).
Anything maintained is fine by me, I mean it, I recode everything  
needed
in my software, but methods are needed for the average user.
If they are existing I apologize, otherwise I just say let's add  
them to
"annotate", "AnnotationDbi", or where you think they should sit.
Out of my ignorance, still, I ask:
do we need species metadata packages for what it is in 1 flat  
file in
Homologene?
Since it is ignorance, be nice.
Luigi


On Nov 6, 2009, at 2:41 AM, Tony Chiang wrote:

Hi Luigi,

You might want to also have a look at the homologue annotation  
packages
that can be found in Bioc. They are based up imparanoid. For  
instance the
package for human would be

hom.Hs.imp.db

Cheers,
--Tony

On Thu, Nov 5, 2009 at 11:10 PM, Luigi Marchionni  
<marchion at jhu.edu>wrote:

Dear All,
As I  wrote to the list a couple of weeks ago I took on the  
endeavor of
creating an S4 package for storing genomics results data and  
further analyze
them.
I had already code working to compare results across experiments,
platform and species.
To be a good citizen I start using S4, and I start relying on  
all classes
already existing in Bioc.
Now I came to the issue of dealing with mapping genes (and  
features)
across species.
I see that Hong Li maintains a package (homolog.db) containing  
such
information, which depends on several other packages.
I installed them and found difficult to use it.
I will give you few examples:

This retrieves the mapping between the Homologene ID and the  
Entrez Gene
ID.
Obviously each list element has a different length, however  
there is not
easy way to tell the correspondence between organism and Entrez  
gene ID.
I can say that the first 1 in both elements below is Human,  
then...
If this has to be the structure, then each element in xx below  
should be
names with the corresponding taxonomy id.
See the chunk of code below:


################################################################################

xx <- as.list(homologHOMOLOG2GENEID)
xx[1]

$`3`
[1]      34  469356  490207  505968   11364   24158  406283
[8]   38864 1276346  181757  173979  181758

################################################################################

By using the code below I can however retrieve the mapping  
between Entrez
gene identifiers to Homologene identifiers.
Lets consider the first two elements of xx[1] above:


################################################################################

yy <- as.list(homologHOMOLOG)
yy["34"]

$`34`
[1] 3

yy["469356"]

$`469356`
[1] 3

################################################################################

Using a little coding I can now map from one Entrez ID to  
another across
species, although without knowing which species. So I can use  
species
information:


################################################################################

zz["34"]

$`34`
[1] 9606

zz["469356"]

$`469356`
[1] 9598

################################################################################

OK. now I know that Entrez ID "34" in Taxonomy "9006" (human)  
correspond
to Entrez ID "469356" in n Taxonomy "9598" (which I do not know  
by heart),
through the Homologene id "3". To learn the the second taxonomy  
I can do:


################################################################################

ff <- as.list(homologORGANISM)
ff["9598"]

$`9598`
[1] "Pan troglodytes"

################################################################################

Good!  I had to play around a little with the code, however I  
could map
the human Entrez ID "34" to the monkey "469356" one.
However I think this is a little too complicated. To install  
homolog.db
and (with dependencies=TRUE) I also had to install:
org.Hs.ipi.db_1.1.1.tar.gz
org.Hs.sp.db_1.1.1.tar.gz
PAnnBuilder_1.9.0.tar.gz
And the package does not point to a library that implements the  
chunks of
code above to map Entrez ids across species.

Look the code below, I load my mapping library (where the cross- 
mapping
homologene table takes 3.2 Mb), I load this object, and the  
taxonomy
information:


################################################################################

library(moreFGS)
data(homol)
data(tax)
ls()

[1] "ff"    "homol" "tax"   "xx"    "yy"    "zz"

################################################################################

Finally I load a library containing the  taxSwitch() function:


################################################################################

library(funcBox)
args(taxSwitch)

function (IDs, org1, org2, whatIn = "EGID", whatOut = "EGID")
NULL

################################################################################

Now look at this, for one ID:


################################################################################

taxSwitch("34","Homo","Pan","EGID","EGID")

[1] "469356"

taxSwitch("34","Homo","Pan","EGID","EGID")

[1] "469356"

taxSwitch("469356","Pan","Homo","EGID","EGID")

[1] "34"

taxSwitch("469356","Pan","Homo","EGID","symbol")

[1] "ACADM"

taxSwitch("34","Homo","Mus","EGID","symbol")

[1] "Acadm"

taxSwitch("Acadm","Mus","Homo","symbol","EGID")

[1] "34"

taxSwitch("Acadm","Mus","Pan","symbol","EGID")

[1] "469356"

taxSwitch("Acadm","Mus","Bos","symbol","EGID")

[1] "505968"

taxSwitch("Acadm","Mus","Bos","symbol","Acc")

[1] "NP_001068703"

taxSwitch("NP_001068703","Bos","Rattus","Acc","symbol")

[1] "Acadm"

################################################################################

Or more than one ID:


################################################################################

taxSwitch(c("34","37","3211"),"Homo","Mus","EGID","Acc")

[1] "NP_031408" "NP_059062" "NP_032292"

taxSwitch(c("34","37","3211"),"Homo","Mus","EGID","symbol")

[1] "Acadm"  "Acadvl" "Hoxb1"

################################################################################

and so on.
I would be very happy to provide bioconductor with the code to  
make the
moreFGS library and with the taxSwitch() function.

Luigi

PS: the session info is below


################################################################################

sessionInfo()

R version 2.11.0 Under development (unstable) (2009-10-01 r49916)
i386-apple-darwin9.8.0

locale:
[1] en_US.UTF-8/en_US.UTF-8/C/C/en_US.UTF-8/en_US.UTF-8

attached base packages:
[1] stats     graphics  grDevices utils     datasets
[6] methods   base

other attached packages:
[1] moreFGS_1.0.2       homolog.db_1.1.1
[3] PAnnBuilder_1.9.0   RSQLite_0.7-3
[5] DBI_0.2-4           funcBox_0.0.3
[7] annotate_1.25.0     AnnotationDbi_1.9.0
[9] Biobase_2.7.0       limma_3.3.1

loaded via a namespace (and not attached):
[1] tools_2.11.0 xtable_1.5-5

################################################################################

	[[alternative HTML version deleted]]