Re: [Obo-discuss] develops_from

[Melissa H. <ha...@oh...>]

Hi all,
few comments.

On Oct 28, 2010, at 2:45 AM, David Osumi-Sutherland wrote:

Hi all,

On 28 Oct 2010, at 28/Oct/2010 03:02:46, Chris Mungall wrote:

In fact this is mostly of theoretical interest, as to my knowledge
none of the anatomical ontologies ever adopted "derives_from" - as
Pankaj points out, the label is misleading (evo folks are particularly
unhappy), and distinguishing transformation_of from derives_from was
considered too difficult.

Please let's keep the evolutionary usage distinct.
completely agree.

In addition, very many existing
develops_from relationships are neither transformations nor
derivations, because the parent tissue continues to exist. Instead
everyone opted to keep the fuzzy but familiar "develops_from", and the
RO definition of derives_from was relegated to the ivory tower and
never applied.

I think the way forward as far as most bio-ontologies are concerned is
to focus on biological definitions.

Melissa summarizing David's talk from the cell meeting:

At the cell meeting we discussed the idea that cells can derive_from
one another as defined in RO, where a derivation would involve some
kind of cell division and a transformation would not. A neuroblast
which does not undergo a cell division an becomes a neuron would be
a transformation. A neuroblast that buds off to give rise to a
neuron would be a derivation.

Here's what I suggested at the cell-type meeting as suitable relations for cell lineage (the hierarchy here is meant to be a property hierarchy, so arises_directly_from SubPropertyOf arises_from; please don't worry too much about the names)

•arises_from (transitive)
–arises_directly_from (non-transitive)
•buds_from (cell maintains its identity (and its class) while daughter cell buds from it).
–e.g.- ganglion mother cell buds_from some neuroblast
•fusion_of (two or more cells fuse to form one multi-nucleate cell; one or both cells change identity (and class) in the process)
–e.g.- muscle fibre development by fusion of myoblasts
-(We can’t use ternary, perhaps multiple binary sufficient?)
•fission_of (cell divides, daughter cells have different identities (and classes) to parent)
–e.g.- ganglion mother cell divides to form one neuron and one glial cell
•transformation_of (cell differentiates => new class (but not new identity?). No cell division involved.)
–e.g.- T-cell transforms to activiated T cell (?)

I've only defined the leaf node relations here. I'd be happy to define the parent terms as unions.  Also, I think that the reason these are reasonably straightforward to define is that it is clear what the relevant level of granularity is here: the cell level.  This means that we don't have to worry that exocytosis is budding or that endocytosis is fusion.  I'm also aware that I've glossed class vs identity. Now, from discussion here and re-reading the OBO relations paper, is is clear to me that arises_directly_from can not be the same as derives_from, because it encompasses budding.  If I understand correctly, this is because of the problem of identity that I've glossed.  Nevertheless, I still think that my suggestion is a useful and practical way to model cell lineage that mirror pretty well the way biologists think about it. I'd rather not hold up implementing it, or something like it because we get bogged down in thorny philosophical issues around the problem of identity or because we can't agree on more abstract relations that apply at every level of granularity from molecules to stars.
Yes we've been here many times before. Am I the same person as yesterday? I cut my fingernails ;-)


Where this breaks down is on the gross anatomical level, where the
level granularity is arbitrary. When developmental biologists say
one gross structure develops from another, they mean that there is
continuity of cell lineage.

I'm confident from this we have develops_from between cells covered -
though we have to think a bit about the relation between the cels in a
cell line and the in vivo progenitor.

As Melissa says the gross level is harder. We can define a chain:

A has_part cell C
C1 develops_from cell C2
C2 part_of B
==>
A develops_from B

But this is almost certainly too permissive.

I'd say that this is a necessary condition but that it is not sufficient. It is definitely too permissive - certainly for early vertebrate development where fate maps are probabilistic and there are cases where cells have achieved some degree of specification, but are not yet determined (they might be re-programmed depending on the signals present in the micro-environment where they end up.)


We might need to fudge it
with something hard to formalize like "X% of cells", or perhaps it can
be done using the definition of tissue.

Perhaps. But if 'develops_from o part_of -> develops_from', then we can't even specify a majority of cells.

It may be that the only practical way to proceed is to just specify some necessary conditions for the use of develops_from (or some such relation) between multi-cellular structures.   Some restrictions on usage would be an improvement over the current situation.
Agree here too. Currently, the develops_from relation is being used across many ontologies, and just having necessary conditions specified would a great improvement.



- David


It would be good for some of the dev bio folks to come up with some
edge cases we can explore.




David Osumi-Sutherland, PhD
Curator/ Ontologist
FlyBase / Virtual Fly Brain
Department of Genetics,
University of Cambridge,
Downing Street,
Cambridge, CB2 3EH, UK
Tel: +44 (0)1223 333 963
Fax: +44 (0)1223 766 732

<ATT00001..txt><ATT00002..txt>

Dr. Melissa Haendel

eagle-i Networking Research Resources
Oregon Health & Science University
ha...@oh...<mailto:ha...@oh...>
skype: melissa.haendel
503-407-5970