[Bioc-devel] Multiple colData in SummarizedExperiment

you can just implement this by having reserved column names in the colData
slot; that will work and will take appr. 23 seconds to implement.  I agree
it is not as clean from a design perspective, but you get 100% of the
functionality and you can write a separate checker for the colData argument.

On Thu, Jun 18, 2015 at 2:00 PM, davide risso <risso.davide at gmail.com>
wrote:

Thank you all for the responses.

I didn't think about the nested DataFrame solution.  It should work.
I agree that an extension might be cleaner, but I clearly need to give it
more thought.

One of the reasons I wanted to have quality and metadata as separate
slots is that one could enforce that all the qualities are numeric, and
have a quality() method to extract just the quality scores (e.g., for
plotting / quality control). Having them in the same slot makes it harder
for the user to extract just the scores (if the column order and/or names
are not standardized).

Best,
davide


On Thu, Jun 18, 2015 at 6:35 AM Vincent Carey <stvjc at channing.harvard.edu>
wrote:

yes, if a formal extension is warranted.  the metadata slot could also be
used.

On Thu, Jun 18, 2015 at 2:59 PM, Kasper Daniel Hansen <
kasperdanielhansen at gmail.com> wrote:

I think the more clean solution for Davide (if he inists on having

separate

objects; I decided against it in minfi) is to extend the class to allow
this.

Kasper

On Thu, Jun 18, 2015 at 12:25 AM, Ryan <rct at thompsonclan.org> wrote:

Oh wow, I didn't know you could put a DataFrame into a single column

of

another DataFrame. That actually solves a problem for me too (I don't
intend to expose nested DataFrames to the users though).


On 6/17/15 7:23 PM, Martin Morgan wrote:

On 06/17/2015 11:41 AM, davide risso wrote:

Dear list,

I'm creating an R package to store RNA-seq data of a somewhat large
project
in which I'm involved.

One of the initial goals is to compare different pre-processing
pipelines,
hence I have multiple expression matrices corresponding to the same
samples.
The SummarizedExperiment class seems a good candidate, since I have
multiple expression matrices with the same rowData and colData
information.

I have several sample-specific variables that I want to store with

the

object, namely, experimental information (e.g., batch, date,

experimental

condition, ...) and sample quality (e.g., proportion of aligned

reads,

total duplicate reads, etc...).

Of course, I can always create one big data frame concatenating

the two

(experimental info + sample quality), but it seems that both

conceptually

and practically, it might be useful to have two separate data

frames.

Since this seems somewhat a reasonably standard type of information

that

one would want to carry on, I was wondering if it would be

possible /

useful to allow the user to have multiple data.frames in the

colData

slot

Actually, colData() is a DataFrame, and a DataFrame column can

contain a

DataFrame. So after

  example(SummarizedExperiment)

we could make some faux sample quality data

  quality = DataFrame(x=1:6, y=6:1, row.names=colnames(se1))

add this as a column in the colData()

  colData(se1)$quality = quality

(or create the SummarizedExperiment from a similar DataFrame

up-front)

and manage our grouped data

colData(se1)

DataFrame with 6 rows and 2 columns
    Treatment     quality
  <character> <DataFrame>
A        ChIP    ########
B       Input    ########
C        ChIP    ########
D       Input    ########
E        ChIP    ########
F       Input    ########

colData(se1[,1:2])$quality

DataFrame with 2 rows and 2 columns
          x         y
  <integer> <integer>
A         1         6
B         2         5

I'm not sure that this is any less confusing to the end user than

having

to manage a DataFrameList(), but it does not require any new

features.

Martin

 of SummarizedExperiment.

Best,
Davide

    [[alternative HTML version deleted]]

        [[alternative HTML version deleted]]

        [[alternative HTML version deleted]]