FirebrowseR

[Viktoria Kliewer]

Dear Sir and Madam,

I am an assistant at the University of Technology Dortmund, Germany, at
the department of Computer Science VIII. Sangkyun Lee, a research associate
at the department of Computer Science VIII, and me, we use FirebrowseR,
especially the services Samples.ClinicalTier1, Analyses.Mutation.MAF,
Samples.miRSeq and Samples.mRNASeq, to achieve the following objectives:
For about 8 months we have been experimenting with FirebrowseR and
developing algorithms to download the whole available particular TCGA data
for every cancer respectively. After developing the algorithms we aim for
restructuring the data in the form of matrices so that it could be
possible to find patterns in these data.

Actually our algorithms are implemented. But it appeared a problem.
All services we use require at least one gene/miRNA_ID/mRNA_ID or barcode
to return an output. So there are to ways specifying either genes or
barcodes:
1. specify barcode(s), gene="":
service(tcga_participant_barcode, gene) -> output
2. specify gene(s), tcga_participant_barcode="":
service(tcga_participant_barcode, gene) -> output

To get all available data of a cohort we can use two possibilities:
a. parallelize over barcodes
b. parallelize over genes

The possibility b. is more efficient and takes less time.
Two weeks ago we determined that the first way,
service(tcga_participant_barcode) -> output, doesn´t work anymore. The
output is NULL regardless to the input barcodes.
Now there is a dilemma: the possibility b. works, but to get all the
genes/miRNA_IDs/mRNA_IDs it is nesessary to use the first way.
So our question: Has something been really changed in the server side?

I hope that I illustrated our problem correctly and understandably.
As we currently can't move any further without a feedback from you, we
would very much appreciate a timely answer.
Thank you a lot in advance.

Kind regards,
Viktoria Kliewer

[Michael Noble]

Dear Viktoria,

Thank you for writing such a detailed note.  I’m glad to hear that FireBrowse has been helpful in your research.  In general we’d be happy to assist your design planning, but please note, however, that we do not maintain the R bindings for FireBrowse here at the Broad Institute.  Rather, they were developed by Mario Deng et al (cc:d), so let’s include him (and his colleagues) into the discussion and see what the most effective approach might be to move forward.

Best, 

-Mike

—

Michael S. Noble

Associate Director for Data Science

Cancer Genome Computational Analysis

The Broad Institute of MIT & Harvard

Manager, Genome Data Analysis Center

The Cancer Genome Atlas

[Michael Noble]

No problem, Viktoria.  To one of your original questions, though:  yes, we recently did change something in the server API for querying by barcodes.  This was mentioned in the Release Notes (see v1.1.30, section 3.e and 3.f).  We did this to avoid a database problem, where barcode-only queries of the mRNASeq and mIRSeq data would consume a great deal of time and compute resources.  We haven’t had enough time to provide similar feature in a new API yet, but we will very soon—I hope by July.  I’m sorry this has caused a problem in your analytics, but I appreciate that you took the time to mention it to us because it helps us know what is important to users (and thus where to spend our limited time).

Best,

Michael

On Jun 6, 2016, at 11:53 AM, Viktoria Kliewer <viktoria...@tu-dortmund.de> wrote:

Dear Mike,

thank you a lot for such a timely answer and for being cooperative!
Yes, FireBrowse is very helpful, it´s the key component in our algorithms.
As Google didn´t tell me who is the developer of FirebrowseR and I didn´t
found any other email address I contacted you. So thank you for refering
to the right contact persons.

PS: I noticed a mistake in my mail: the possibility a. is more efficient
and takes less time, not b.

Best regards,
Viktoria

Dear Viktoria,

Thank you for writing such a detailed note.  I?m glad to hear that
FireBrowse has been helpful in your research.  In general we?d be happy to

assist your design planning, but please note, however, that we do not
maintain the R bindings for FireBrowse here at the Broad Institute.

Rather, they were developed by Mario Deng et al (cc:d), so let?s include

him (and his colleagues) into the discussion and see what the most
effective approach might be to move forward.

Best,
-Mike

?

['Mario Deng' via gdac]

Dear Viktoria,

the data provided by Firebrowse are made available thru an web-API [1] and the FirebrowseR [2] package retrieves its data from this API.

Both, the API as well as the corresponding bindings (the R FirebrowseR Package) are update constantly. So, my first question would be: Are you using the current version of FirebrowseR? The current API-Version is 1.1.32 (as is for FirebrowseR), to check which version of FirebrowseR is currently installed run

require(FirebrowseR)

sessionInfo()

from your R session. To me it seems like you are using an outdated version, since the function "Samples.ClinicalTier1" is just not available anymore.

Regarding your initial question I am a bit stumped, since I don't know what your "service" function is doing. Also it is hard to provide any feedback, since I can't reproduce your errors. You could provide us with an working example [3] to respond to your specific problem with FirebroweR.

Best,

MD

[1] http://firebrowse.org/api-docs

[2] https://github.com/mariodeng/FirebrowseR

[3] http://stackoverflow.com/help/mcve