RE: [Mged-mage] mage discussion archives and package relationship s

[Angel P. <an...@sn...>]

On Mon, 6 Jan 2003, White, Joseph A. wrote:

> Hi Angel,
>
> OK, I misread the UML.  What I'm trying to discern is what CAN go into the
> BioAssayDimension of BioAssayData.  If any one BioAssay can only have one
> BioDataCube, what's the BioAssayDimension for?

Any BioAssay Can have more than one BioAssayData and hence more than one
BioDataCube. Also any BioAssayData object can have more than one BioAssay,
hence the BioAssayDimension needs to state the ordering of these assays. I
think that you are equating the BioAssayData object with the BioDataCube.
A BioDataCube is justa  3D matrix. It needs the descriptive info from MBAD
or DBAD objects in order to make sense of the numbers.


> Obviously, it is to identify the BioAssay, but if it's a dimension of
> 1, why wasn't this treated as a simple matrix?

It can be, but you must always account for a 3D matrix. So the
BioAssayDimension has only one component, in your example.


>
> So here's the scenario:
>
> hybridization        scan                    feature extraction
> -------------        ----                    ------------------
> PhysicalBioAssay --> ImageAcquisition(1) --> MeasuredBioAssay(1) -->
> BioAssayData(1) --> Cube(1)
>                                          --> MeasuredBioAssay(2) -->
> BioAssayData(1) --> Cube(???)
>                  --> ImageAcquisition(2) --> MeasuredBioAssay(3) -->
> BioAssayData(2) --> Cube(2)
>
> (Yes, we have had people scan slides more than once.)
>
> So, the bottom line question is: CAN the data from MBA(2) go into
> BioDataCube(1), or do they have to go into a new cube.

It can go into the same cube, since you have two MBA objects as the
BioAssayDimension components.

> (Certainly the data from the second ImageAcquisition will go into a
> new cube.)

Well, strictly speaking, if the ImageAcq(2) outputs the same parameters as
ImageAcq(1), then even this can go into the same cube. You can determine
which values came from which scan by backtracking through the MBA object.

I personally would do it the same way that you describe, but you have to
account for both in your import scripts.

> The UML indicates that each of these MBAs must go into a
> separate cube--or am I reading this incorrectly?  If that is correct,
> what other BioAssays can be placed in the BioAssayDimension of
> BioAssayData ?

As long as the *Dimension objects accurately describe all the axis of a
given slice through the data, then you can lump any sort of data together
in the same cube.

The reason that DerivedBioAssayData usually ends up in their own cube is
due to the numerous transformations done on the matrix that produce the
data in the first place. In other words, the very act of creating a
DerivedBioAssayData instance will usually negate either the
BioAssayDimension (avg across BioAssays), DesignElementDimension (avg.
across duplicate features/reporters), or QuantitationTypeDimension (adding
a ratio column). Hence the need for a new cube a *Dimension descriptions.


Angel