obo-relations
obo-relations copied to clipboard
oborel
participates in -- what sorts of sub-relations should we identify?
first ideas:
is agent of is patient of is principal participant of
At GO we require more specification of participants in biochemical reactions and processes. We need to be able to distinguish between the enzymes that catalyze reactions, the primary molecules that are the focus of a biosynthetic or catabolic pathway, cofactors, substrates, intermediates etc.
(Note there is some previous discussion here #18, but that ticket looks ugly as it was imported from googlecode, I will close that one and update discussion here, see also #171 )
Historically we have divided reaction participation relations into 3 roles:
- input
- output
- enabler/catalyst
- RO:0000057 ! has participant
- RO:0002233 ! has input
- RO:0002234 ! has output
- RO:0002333 ! enabled by
NOTE: I suggest we restrict this ticket to discussion of input/output, otherwise the discussion will get confusing. Open a new ticket for agent/enabler or add to #171
input/output are both subtypes of "patient" roles in Barry's comment (we can use this terminology here amongst us for discussion, but this terminology should never be used in RO since everyone will assume "patient" has the lay interpretation).
This is sufficient for a biochemical database but when we want to represent processes tha t are the realizations of evolved functions we need more specific relations to be able to separate the "intended" inputs and outputs from cofactors or byproducts. This may map to Barry's 'principal' participant.
We have a separate ticket open for a 'has substrate' subproperty of participates in #230
Proposed schema for evolved processes. Evolved processes are either (1) the realization of a single function (2) a causally connected mereological sum of (1)s
The schema could be generalized to other processes arise from non-evolved functions such as building a house. For simplicity this initial elucidation focuses on evolved processes (aka biological programs).
Subsubmption hierarchy (P = process, C = independent continuant, Q = quality)
- enabled by (aka has agent)
P->C - has participant
P->C- modifies or consumes (aka has initial participant)
- has input (aka has primary input)
- has substrate
- ...
- has secondary input
- has input (aka has primary input)
- produces (aka has terminal participant)
- has output (aka has primary output)
- has secondary output
- has intermediate
- modifies or consumes (aka has initial participant)
- process has state
P->Q - process has initial state
- process has input state
- ...
- process has terminal state
- process has output state
Additional grouping relation not shown: acts upon
Notes
- Note that has agent is not a sub-relation of has-participant. It is possible for the agent to "sit outside" or overlap the process.
- I propose reserving the names "has input" and "has output" for a sense that is more restricted than some may be comfortable with
Definitions
has input
def: A relation between an evolved process P and an independent continuant C such that C participates in P and the objective of P is to change C, either to destroy/consume it or to change one of more of the qualities inhering in C
Examples:
- a ubiquitinated protein in a protein ubiquitination process
- an embryo inside of a gravid organism in a birth process
has output
def: A relation between an evolved process P and an independent continuant C such that C participates in P and the objective of P is to bring into existence a C
Examples:
- a deubiquitinated protein in a protein ubiquitination process
- an neonate outside of a gravid organism in a birth process
modifies or consume
def: A relation between an evolved process P and an independent continuant C such that C participates in P and C is present at the start of P, and either the state of C changes during P or C is consumed/destroyed
has secondary output
def: A relation between an evolved process P and an independent continuant C such that C modifies or consumes P and C is not the primary output of P
Examples:
- a ubiquitin molecule in a protein ubiquitination process
- afterbirth
acts upon
def: P acts upon C iff P has_input C and P has_output C
(trivially: the input and output are the same entity, albeit in a different state)
enabled by (aka has-agent)
elucidation: a relationship between P and C such that C directly causes P to happen.
Note: not every evolved process has an agent. Some processes will be the sum of multiple processes each with an agent. For example, every step in a signaling pathway can be assigned an agent but the overall program need not have an agent.
TO BE DECIDED:
we may want to give this a definition of: P enabled by C iff C bearer of some F and F realizes P, but does this entail that C is a participant? We want to reserve the right of the agent to be outside the process. If this is entailed, then the proposed def can be modified to include direct regulation.
- an enzyme or macromolecular complex in a molecular reaction
- a person driving a car
State/Quality Relations
These follow the same pattern as the `C relations above but allow us to be more precise, in stating the state of affairs at the start and end.
In fact the C relations should probably be defined in terms of the Q relations.
E.g.
P $qrel Q, Q inheres_in C => P $crel C
has input
def: A relation between an evolved process P and an independent continuant C such that C participates in P and the objective of P is to bring into existence a C
Looks like input and output are switched here? Surely an input must exist before the process.
On 1 Aug 2018, at 16:02, Chris Mungall [email protected] wrote:
Note that has agent is not a sub-relation of has-participant. It is possible for the agent to "sit outside" or overlap the process.
(overlap sounds wrong here - if mereology should be P to P or C to C not P to C)
It seems like these are meant to be the examples:
an enzyme or macromolecular complex in a molecular reaction a person driving a car
Does a driver really not participate in driving? It seems odd to say that an enzyme doesn’t participate in the process it catalyses given that most bind to their substrates.
If these agents don’t participate in the process they enable are they participating in a process that regulates it?
I don’t want to be dogmatic though. I think this should be decided on the basis of: (a) what is most intuitive to users (restricted to biologists?) and (b) what best supports use-case queries (e.g. for what use cases is it important to keep agents distinct from participants)
Note: not every evolved process has an agent. Some processes will be the sum of multiple processes each with an agent. For example, every step in a signaling pathway can be assigned an agent but the overall program need not have an agent.
What is the relationship of the overall process to the agents that carry out its various steps? Are they not agents of that process?
Looks like input and output are switched here? Surely an input must exist before the process.
You're right. I edited my original comment to fix this.
(overlap sounds wrong here - if mereology should be P to P or C to C not P to C)
Correct. I lapsed into 4D-ism
I didn't mean to imply that the agent never participates, just that we may need to be flexible. But also open to being consistent and always considering the agent a participant by definition.
What is the relationship of the overall process to the agents that carry out its various steps? Are > they not agents of that process?
I think it is useful to restrict it this way, but there's nothing to stop up naming a relation has_part_with_agent <- has_part o agent (including the reflexive case, possibly with a friendlier name)
I have the same concerns about excluding agents from participants; it seems counter-intuitive to say that gene products that enable a process (single-step function) are not participants in that process.
I am also wondering about the 'primary' vs 'secondary' input distinction and whether that distinction is preferred over just having more specific children of has_input.
Using protein ubiquitination as an example, could we describe the input ubiquitin with a relation like 'has_donor' and the protein that is ubiquitinated as 'has_acceptor' both of which would be children of has_input, or possibly even children of has_substrate?
OK, everyone seems to be in favor of inclusion for agent
Specific relations: this is certainly less ambiguous. And for those of us working with a tbox (ontology) we can take advantage of standard documentation in design patterns.
Some have objections to what are perceived as proliferating relations. These are largely mitigated if certain criteria are met (definition of relations, probably via a role hierarchy, documentation in DPs). Nevertheless it does incur additional costs in the form of cognitive burden of curators doing post-composition. It's not always straightforward to get around this with tooling.
On Wed, Aug 1, 2018 at 7:02 PM, Chris Mungall [email protected] wrote:
Proposed schema for evolved processes. Evolved processes are either (1) the realization of a single function (2) a causally connected mereological sum of (1)
The schema could be generalized to other processes arise from non-evolved functions such as building a house. For simplicity this initial elucidation focuses on evolved processes (aka biological programs).
Can we then find a replacement for 'Evolved' , since there are functions not only in the biological world but also in the world of artifacts (such as pacemakers, syringes, ...). And should we be allowing also realizations of capabilities to count under (1). (E.g. you have the capability to digest food, but it is not your function to digest food.)
Subsubmption hierarchy (P = process, C = independent continuant, Q =
quality)
- enabled by (aka has agent) P->C
I insert here your definition of 'enabled by'?
enabled by (aka has-agent)
elucidation: a relationship between P and C such that C directly causes P to happen.
Is there a relation 'causes to happen', with 'directly causes to happen' as a child?
If so, can we have at least an elucidation of the former and (ideally) a definition of the latter?
- has participant P->C
- modifies or consumes (aka has initial participant)
- has input (aka has primary input)
how can 'has input' be aka 'has primary input' if there is also such a think as secondary input
- has substrate
- ...
- has secondary input
- produces (aka has terminal participant)
'terminal' sounds risky; do you have a definition?
- has output (aka has primary output)
again, this aka looks wrong
- has secondary output
- has intermediate
intermediate what?
- process has state P->Q
what is your definition of 'state'?
- process has initial state
- process has input state
- ...
- process has terminal state
- process has output state
Additional grouping relation not shown: acts upon Notes
- Note that has agent is not a sub-relation of has-participant. It is possible for the agent to "sit outside" or overlap the process.
Example?
- I propose reserving the names "has input" and "has output" for a sense that is more restricted than some may be comfortable with
This sounds *very *risky, especially given that the OBI
treatment of these terms is now being used by many people outside biology. The foundations of RO should be at the same level of generality as BFO, if it is to work across all the communities who will want to use it.
Definitions has input
def: A relation between an evolved process P and an independent continuant C such that C participates in P and the objective of P is to bring into existence a C
Examples:
- a ubiquitinated protein in a protein ubiquitination http://purl.obolibrary.org/obo/GO_0016567 process
- an embryo inside of a gravid organism in a birth process
can you define 'objective'?
has output
def: A relation between an evolved process P and an independent continuant C such that C participates in P and the objective of P is to change C, either to destroy/consume it or to change one of more of the qualities inhering in C
can an inspection process have an output (other than the inspection report)?
Examples:
- a deubiquitinated protein in a protein ubiquitination http://purl.obolibrary.org/obo/GO_0016567 process
- an neonate outside of a gravid organism in a birth process
modifies or consume
def: A relation between an evolved process P and an independent continuant C such that C participates in P and C is present at the start of P, and either the state of C changes during P or C is consumed/destroyed
has secondary output
def: A relation between an evolved process P and an independent continuant C such that C modifies or consumes P and C is not the primary output of P
It is very confusing to have 'has secondary output' not be a child of 'has output'. I believe you should avoid 'aka', and just have one term for each relation, at least for our purposes here.
Examples:
- a ubiquitin molecule in a protein ubiquitination http://purl.obolibrary.org/obo/GO_0016567 process
- afterbirth
acts upon
def: P acts upon C iff P has_input C and P has_output C
(trivially: the input and output are the same entity, albeit in a different
state)
destroying/consuming is not acting upon?
enabled by (aka has-agent)
elucidation: a relationship between P and C such that C directly causes P to happen.
Note: not every evolved process has an agent. Some processes will be the sum of multiple processes each with an agent. For example, every step in a signaling pathway can be assigned an agent but the overall program need not have an agent.
TO BE DECIDED:
we may want to give this a definition of: P enabled by C iff C bearer of some F and F realizes P, but does this entail that C is a participant?
This would mean that we do not need to define 'cause' after all, which may be a good thing.
We want to reserve the right of the agent to be outside the process.
Still good to have examples of what you mean by this.
If this is entailed, then the proposed def can be modified to include direct regulation.
- an enzyme or macromolecular complex in a molecular reaction
- a person driving a car
State/Quality Relations
These follow the same pattern as the `C relations above but allow us to be more precise, in stating the state of affairs at the start and end.
'state' standardly means something very different from 'state of affairs' (the latter is depicted by a true sentence, the former by something like a quality expression [e.g. on, off, pregnant, ...]
In fact the C relations should probably be defined in terms of the Q relations.
E.g.
- P $qrel Q, Q inheres_in C => P $crel C
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/oborel/obo-relations/issues/244#issuecomment-409753107, or mute the thread https://github.com/notifications/unsubscribe-auth/AH6qNzmko1Y4QwwJNc2UVOVggFph3Qndks5uMjOWgaJpZM4VigR6 .
I think we need to ensure we allow a general participation relation. For example, I am a person that participates_in ICBO2018. I could have multiple roles that are realized in this context: 'workshop organizer role', 'presenter role', etc. I think this ICBO2018 may not be an an evolved process, and doesn't fit include either (1) the realization of a single function, or (2) a causally connected mereological sum of (1)s as the roles being realized are not causally connected. Is this still adherent under the current thinking?
Just want to make sure we don't lose the ability to use the participates_in relation in a general way, as it is often used for social structures and other contexts.
One possible solution to the problem of not being able to come up with a good definition for "enables" (or, relatedly, "agent") is to define its inverse. See https://github.com/geneontology/go-ontology/issues/16264 which discusses a definition of a currency metabolite based on frequency of occurrence in different reactions. "Enabling" inputs are those that are not currency metabolites.
