Re: [Gusdev-gusdev] schemas

[Arnaud K. <ax...@sa...>]

Hi Chris

Sounds good. See below for some comments.

Chris Stoeckert wrote:

> Hi All,
> Here are the phenotype tables I mentioned during the conference call. 
> I still need to look at the rest of the schema to see how best to link 
> it to gene sequences but welcome suggestions.


We would use it to annotate alleles. I guess a new table is needed with 
the allele name, the degree of dominance, mutation data, complementation 
data ...

>
> Also, Angel has posted a picture of a proposed schema for capturing 
> mass spec data at:  
> http://www.cbil.upenn.edu/downloads/Proteomics/schema/
> Next steps are to have people look it over and try to fit available 
> data into it.
>
> Finally, I will write up my notes of our conference call and send to 
> Marie-Adele to edit and pass around (if she agrees ;-) ).
>
> Cheers,
> Chris
>
> Phenotype: A linking table in the spirit of GOBO or PATO constructing 
> phenotypic descriptions from orthogonal ontologies using nounal phrases.
> phenotype_id
> pato_attribute_id
> anatomy_id
> cell_type_id        can point to the Anatomy table
> dev_stage_id
> description    varchar 255    text description of what the terms 
> represent in aggregate


At the moment, as pombe, Trypanosoma and Leish are unicellular, the 
orthologous ontologies we would use are GO process and GO component, as 
well as the Parasite life cycle ontology being developed here by Mat. 
Berriman.

>
> PATOAttribute: A
> pato_attribute_id
> term_name    varchar255
> term_synonym    varchar 255
> version        varchar 50
> parent_pato_attribute_id       term_type    varchar 20    "attribute", 
> "attribute value"

So to summarize a phenotype is made from a set of ontology terms, there 
is a many to many relationship between each of these orthogonal 
ontologies and the Phenotype table. It also has a set of attributes, and 
each attribute is associated to a set of attribute values.

cheers
Arnaud


Note this is based on Michael Ashburner's draft.
----------
On the representation of phenotypic data in genome databases - 
pato.ontology.

A discussion paper.

Version 2.0 22 February 2002.

Michael Ashburner, FlyBase.

pato: phenotype & trait ontology.

This paper assumes some knowledge of the work of the Gene Ontology
Consortium - see www.geneontology.org.

A major conceptual and practical problem that faces all genomic 
databases is
how to describe mutant phenotypes (I include in this term epigenetic 
phenotype
due to RNAi, for example, and the more general class of traits, in the 
sense
this word is used by plant and animal breeders).  Some databases are, I 
know,
struggling to find a solution to this problem, these include FlyBase, 
Gramene,
the Mouse Genome Database, and others.  Traditionally, phenotypes have been
described as free text, for example in some FlyBase tables and in the Mouse
Locus Catalog.  This is great, but is obviously a poor solution if the 
data are
to be efficiently analysed (e.g. by a search engine) computationally.  I 
am not
wholly convinced that we can ever get away from free text, at the very 
least as
a "gloss" on a more systematic representation of these data. 
Nevertheless, the
effort to represent these data systematically is very much worth while, 
and is
again - I maintain - an effort that it is far better we do collaboratively,
arriving at common standards, that independently.  If this works it also
has the consequence that rigorous queries across databases could be made
(as now with GO).

The core of my proposal is that phenotypic data can be represented as
qualifications of descriptive nouns or nounal phrases.  A typical 
descriptive
noun would be "eye" or "leaf" or "wing development".  For each class of 
such
nouns there will be a finite (I hope) set of relevant attributes. For these
examples attributes could include: shape, color, size; for "leaf" they 
could
include: thickness, pilosity, for "wing development" they could include
"abnormal". For each attribute there will be a finite (I hope) set of 
values.
For the attribute color, for example, values would include: black, 
white, green.
For each value there will be a finite (I hope) set of qualifiers, for green
these may include dark, light.

Using these three semantic classes we have, I think, the basis for the
systematic  description of phenotypes.  The only extra need is to be 
able to
describe the assay by means of which the phenotypes (i.e. values) were
determined  and the conditions, both environmental and genetic, under 
which the
assay was  performed.  I will not consider assays and their conditions 
in this
paper, since these are very much a concern of the MGED group and we 
should share their work (see www.mged.org).

I see pato as being used to annotate (a) alleles and (b) "strains" or 
"lines"
of an organism (e.g. for crop plants or posssibly mouse strains).

I should say that I do not regard pato as the place for what (albeit 
loosely)
we normally consider as attributes of alleles: e.g recessive, dominant,
cold-sensitive, although pato should be designed to store phenotypic 
data that
is conditional on an environmental condition (or background genotype).

I am also aware that keeping pato in a flat file may be a challenge and 
that a
more sophistical mechanism (?DAML+oil) may be needed.

The purpose of sending this out at this preliminary stage is to get
feedback, both at the conceptual and detailed levels.

I thank John Walshaw, Suzi Lewis, Pankaj Jaiswal, Judy Blake and my 
Cambridge
FlyBase curators (Rachel Drysdale, Gillian Milburn, Chihiro Yamada and 
Aubrey
de Grey) for input & ideas. I have made _no_ attempt to make 
pato.ontology v.1.0
anything like complete - especially with respect to attribute values.

We are experimenting now with re-casting some FlyBase annotation using
pato.ontology, and will report. Please feel free to redistribute this to 
anyone
who might give feedback. Distribution list at end of this file. If you 
never
want to be bothered my me again, send me an email to that effect.

Michael.



The nouns/nounal phrases.
-------------------------

These will be terms from an orthogonal ontology, for example the GO
biological_process ontology, the GO cellular_component ontology or a 
species
specific anatomy ontology (see go/anatomy for the Drosophila and 
Arabidopsis
anatomy ontologies; mouse, worm and Monocot anatomies are on their way).

The attributes.
---------------

The hardest job will be to make a classification of attributes.  A very 
first
step is given below.  We will then need a syntax for declaring for which
nouns/nounal phrases a particular attribute (and its child attributes) 
can be
used.  For exampe attributes of size are of obviously relevant to 
anatomical
structures, but not to a process such as "flight behavior".

Their values.
-------------

The second task is to determine the list of values any attribute may 
have.  I
have shown a few of these in pato.ontology.

Their qualifiers.
-----------------

Finally, we would need to determine the qualifiers that are relevant 
(allowed)
for any particular attribute.  Thus the qualifie fully would be relevant 
to a
fertility attribute, but not to an attribute of size.

I present first a "schema" of the relationships between these concepts:

                                            !antonym             !antonym
                                                |                    |
                                            {mayhave}              {has}
                                                |                    |
concept <-{usedfor}--%attribute ---{has}---> <value ---{has}---> .qualifier
                                           |       |
                                  {determinedby} | {mayhave}
                                         |          |
                                         |          units
                                         |
                                        assay ---{constrainedby}---> 
conditions ---{oftype}---> [MGED]environmental&|genetic


attribute|value|qualifier ---{have}---> synonyms.

concept is a concept (term) from another (orthogonal) ontology.

In the draft pato.ontology: % is for attributes, < is for values of 
attributes