Re: [GeneX-dev] some other data-loading errors found

[<ja...@op...>]

Harry J Mangalam <hj...@ta...> writes:

> PBA (should be renamed PhysicalBioAssay)
>    Def: again, I dunno for sure, but it seems that it refers to the general 
> chip line or product ID or expression technology that is used to quantitate 
> the assay or the organzation that provided the dataset.
> Prerequisites: 
>    Measured BioAssay (optional)
>    Data Provider of type Person or Organization (optional)
>    Array (required)

The BioAssay's are simply collections of annotations associated with
actual data (images are the data for PBA and numerical data matrices
for MBA and DBA).

This is really one of the nicest parts of the MAGE model. Genex-1 only
had the ArrayMeasurement - it sucked, and it sucked well. This was
because we attempted to place all data annotation for data matrices in
a single table - the problem was there were three distinct types of
data that needed annotation: the physical chip data, 'raw' data fresh
off the feature extraction program, and derived data which is
mathematical munging of other data.

MAGE sorts this all out into three types of BioAssay. From the MAGE
Object Model:

  PhysicalBioAssay

  A PhysicalBioAssay is a BioAssay created by the bioAssayCreation
  event (e.g. in gene expression analysis this event is represented by
  the hybridization event).

  MeasuredBioAssay

  A MeasuredBioAssay is the result of direct processing of information
  in a PhysicalBioAssay by the featureExtraction event.  A
  MeasuredBioAssay often uses images that are referenced through the
  PhysicalBioAssay.

  DerivedBioAssay

  A DerivedBioAssay is a BioAssay that is created by the
  Transformation BioEvent from one or more MeasuredBioAssays or
  DerivedBioAssays.

So the idea is:

     1..n   1..n
  PBA -> MBA -> DBA

A single chip (PBA) can produce one or more images (each chip can be
filtered and re-hybridized or simply re-scanned at different PMT
settings), which when run through a feature extraction software
program will produce a 'raw' data matrix (MBA) and when these data
matrices are analyzed they produce one or more derived data matrices
(DBA). 

The Physical BioAssay is all the physical stuff associated with a
hybridization: the chip, the samples, the protocols for hybridization,
washing, and scanning, and finally the images produced are the
Physical BioAssay Data (PBA Data).

The Measured BioAssay is all the meta data associated with the feature
extraction event - it is *not* anything physical, it just records all
the protocols and information about where the MBA data matrix came
from. This is what many consider 'raw' data - the output of the
feature extraction program. MGED and MAGE consider the images to be
the only true 'raw' data (because there is already data processing
that occurs when producing the MBA Data), and so this BioAssay was
called 'Measured' to indicate that it is the 'rawest' possible
numerical data. 

Likewise, the Derived BioAssay is the meta data associated with a data
matrix produced by numerical manipulations of other data matrices
(either MBA Data or DBA Data).

So the BioAssay's are simply ways of assoicating annotations (or meta
data) with the actual 'data' (images for PBA or data matrices for MBA
and DBA). To say it another way, the BioAssay's give you all the
information about how the data was produced - which is seperate from
the data itself - by seperating the annotations from the actual data
it is much easier to do queries because the data matrices are huge
while the meta data is small.

> Array
>    Def: refers to the specific parameters on a per-chip or per-assay basis, 
> including manufacturer, spot concentration, coating, spotter hardware, 
> medium, etc
> Prerequisites: 
>    Spotter Software of type Software (optional)
>    Provider Contact of type Person or Organization (optional)
>    Array Manufacture (required)
>    Array Design (required)
>    Spotter Hardware of type Hardware (optional)
>    Coating (optional)
>    Technology Type (required, but only selectable)
>    Medium (optional, selectable)
>    Default Spot Concentration Units (optional)
>    
>
> ArrayManufacture
>    Def: the date and manufacturers name for backtrack and quality control 
> purposes.
> Prerequisites: 
>    Manufacturer Contact of type Person or Orgazation (optional)

This is a bit of an oddity, but it is the link between the physical
chip and all the LIMS information that went into creating it. So if a
lab wants to track this it is done here.

> Sample (how does this differ from BioMaterial?_

This is a term from Genex-1, if it is used anywhere it is a typo. It
should be Biomaterial.

>    Def: another descriptor for the bioThingie from which you derived the cDNA 
> for the hybridization.  Requires a Data Provider (Person or Org)
> Prerequisites: 
>    Data Provider of type Person or Organization (optional)
>    Taxon (optional)   
>
>
> FactorValues for PBA
>    - ???? (and page comes up with an error)

These are the values for ExperimentalFactor's for any given PBA that
is associated with the Experiment. 

So if an Experiment is a simple time course it has one
ExperimentalFactor, 'time'. Each time point in the time course is
going to have a seperate PBA and each of these will have the
associated FactorValue which is the time point they were harvested. If
an experiment has multiple ExperimentalFactors, so 'time' and 'glucose
concentration', then each PBA may have (but is not required to have) a
FactorValue for each of the ExperimentalFactor's.

In Genex-1 this was called the AM_FactorValues table (in fact MAGE
adopted the genex model for this, but changed the names slightly).

Linking Tables
--

One thing that you've missed (because the insert/edit GUI's can't
currently capture them) are the linking tables in the DB. These are
represented by the apps under the 'Associate' links: 
* 'Groups | Users'          => GroupSecLink table, which controls what
                               users belong to which groups
* 'Roles | Users'           => (this is the Pg system table pg_group)
                               this controls table level privelege,
                               i.e 'ADMIN', 'CURATOR', or 'USER'.
* 'PBA | Protocols'         => ProtocolApplicationLink table, which
                               enables protocols to be associated with
                               a PhysicalBioAssay, e.g. hybridization
                               or washing or scanning (probably wants to
                               be *all* BioAssay's and not just PBA's.
* 'PBA | Biomaterials'      => BiomaterialLink table, allows
                               Biomaterials to be associated with a PBA
* 'Biomaterial | Protocols' => ProtocolApplicationLink table, enables
                               protocols to be associated with
                               biomaterials, e.g. labelling protocol,
                               RNA extraction, cDNA preperation, etc.
* 'Experiment | BioAssays'  => ExperimentLink table, allows BioAssay's
                               to be added to an
                               Experiment. BioAssay's belong to
                               multiple Experiment's.

There are two other linking table that do not have a GUI's:
* QuantitationTypeDimensionLink - this specifies what
  QuantitationType's belong to a QT Dimension - these are entered 
  via the QT Dimension loader and not by hand.
* ManufactureLIMSBiomaterials - this enables Biomaterial's to be
  associated with an ArrayManufacture - so that it is possible to
  identify which nucleic acids are spotted to any given Feature as
  well as an optional link to an external LIMS DB for each Feature
  (just noticed an omission in the schema diagram - this table needs a
  one-to-one link to Feature).

HTH,
jas.