search

obo-behavior / behavior-ontology

Neuro Behaviour Ontology: an ontology for human and animal behaviour processes and behaviour phenotypes
25 stars 15 forks source link

On the relationship of GO:Behavior with NBO:behavior process #101

Open matentzn opened 4 years ago

matentzn commented 4 years ago

So far I went under the assumption that NBO sort of extends the GO behavior branch. For our work in the phenotype ontology world we make extensive use of both, and I need to understand what our framework for reconciling the separately evolving GO:behavior and NBO:behavior process branches will look like.

Is this assumption correct? NBO:behavior process is a subclass of GO:"behavior"

Now, more generally, which of the following is true:

  1. There are NBO terms that are conceptually subclasses of GO terms
  2. There are GO terms that are conceptually subclasses of NBO terms
  3. All of the above.
  4. None of the above.

@cmungall @dosumis @balhoff @pmidford @PaulNSchofield @leechuck

Thanks all!

cmungall commented 4 years ago

I think GO should decide on which of its grandfathered classes truly belong in GO. It should only be there for genetically 'programmed' behaviors. We may have historically over-annotated phenotypes but we shouldn't be doing this any more.

A possibly strange proposal is to conceive of GO and NBO as orthogonal, with GO 'behavior's representing neural circuits/pathways or something at this level, with horizontal edges connecting the two.

matentzn commented 4 years ago

In contrast to your second suggestion, I would like to propose to obsolete NBO:behaviour process and replace it by GO:behavior. From their current definition there does not seem to be that much of a difference:

and

This would make our life forward much easier; NBO needs a bit of a sweep in places, but I would like to defer upper-level alignments to GO, to not having to deal with that ourselves (we will just inherit, for example, GO's alignment with OBOCORE). Also I really need a single root for my patternisation work. :P

matentzn commented 4 years ago

(I did not make any statement yet about orthogonality of GO and NBO and possible MFOEM, for now, just a single upper level alignment that is useful from an engineering perspective and conceptually correct)

matentzn commented 4 years ago

I just saw @cmungall comment GO 'behaviors' are actually neural circuits that shadow (not superclass) the actual behaviors. -- ok so if we dont do what I propose - does this mean that NBO hierarchies and GO behaviors should actually not isa-interleave? So GO behaviours should be used in phenotype definitions only in the context of abnormalBiologicalProcess patterns, and NBO in abnormalBehavior? So they will not really align, unless through some weird shadow relationship?

cmungall commented 4 years ago

does this mean that NBO hierarchies and GO behaviors should actually not isa-interleave?

That is the gist of my proposal, but so far it hasn't gotten any traction, and I admit it sounds unusual. Will need more discussion both on the GO side and the NBO side.

cmungall commented 4 years ago

Note that the GO definition of BP is A biological process represents a specific objective that the organism is genetically programmed to achieve.

If behavior is a subclass of this, then it inherits its properties. It means that all behaviors are genetically programmed objectives. This is a tad problematic. Maybe not for Drosophila. But a wee bit tricky for humans.

For humans, I don't think it's too controversial to say that some behaviors have a genetic basis. But there are some behaviors that may be entirely cultural. And it seems unlikely there are (human) behaviors that are entirely genetically programmed? I suppose you could say that something like sleep is entirely a genetically programmed behavior.

But it's really problematic. The problems go away if we read GO behavior terms not as actual behaviors but as genetically programmed biological processes underpinning behaviors.

I think this split would reflect the different use cases for GO and NBO as well.

matentzn commented 4 years ago

Does this mean that animals without a developed consciousness like worms and flys exhibit only purely genetically programmed behaviors?

It seems that this will result in a big controversy.. Can there be purely genetically programmed behavior? Isnt everything somewhat dependent on environmental factors (including social) but just to different degrees? So you would basically say:

Not sure I can make a lot of sense of drawing a clear line between genetically programmed vs not. But then again, I am not a biologist and defer to @dosumis here :P

dosumis commented 4 years ago

On 12 Feb 2020, at 09:06, Nico Matentzoglu notifications@github.com wrote:

Does this mean that animals without a developed consciousness like worms and flys exhibit only purely genetically programmed behaviors?

It seems that this will result in a big controversy.

Yep. Big can of worms. You’ve tangled consciousness (whatever that may mean) with the ability to learn, which even worms can do to some extent.

Can there be purely genetically programmed behavior? Isnt everything somewhat dependent on environmental factors (including social) but just to different degrees?

We already have this controversy in miniature inside NBO: reflex vs ‘conscious behaviour’ (terrible name for non-reflex in NBO). In a classical view - a reflex is (something like) an innate response to a particular environmental stimulus - i.e. there is an environmental component but the behaviour is not modified by experience. However, this is probably more a matter of degree rather than being categorical - even classical reflexes can be somewhat modified by experience.

So you would basically say:

GO:behaviors: behaviors that occur automatically under normal environmental conditions (availability of oxygen) based on some genetic mechanism (what if the behavior needs a predator to be triggered?) NBO:Behavior with a genetically programmed component NBO:Behavior with an environmentally programmed component NBO: courtship behavior subclass of NBO:Behavior with a genetically programmed component and NBO:Behavior with an environmentally programmed component Not sure I can make a lot of sense of drawing a clear line between genetically programmed vs not. But then again, I am not a biologist and defer to @dosumis https://github.com/dosumis here :P

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/obo-behavior/behavior-ontology/issues/101?email_source=notifications&email_token=AAA3RR3NCIPMFYKPHMPZ7X3RCO32PA5CNFSM4KSNMJWKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOELP7TNI#issuecomment-585103797, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAA3RR5YBWQ5JKR4JF7S643RCO32PANCNFSM4KSNMJWA.

PaulNSchofield commented 4 years ago

A lot of this is about behaviour as phenotype and behaviour as neurological process. So as a phenotype its observable, e.g. nurturing behaviour, psychotic behaviour, goal driven behaviour etc. You can observe it and you can measure it. The classes are therefore operationally defined.

Behavioural process is about neurological processes that result in observable behaviour. Doesn’t imply that all observable behaviour is genetic but that observable behaviour is dependent on the behavioural neurological processes. Maybe a specifically dependent occurrent? I hesitate to get into the philosophy here as there are those on this list better placed to make that call.

As David says, its about the mind/brain problem and thats why its extremely difficult to resolve. We talk about this problem in the NBO papers. Im not sure that Id defend my interpretation for the whole of NBO or GO ( Im sure we can find exceptions to this classification in both) , but I think that it would be interesting to view the domains of each as basically operationally defined phenotypes and underlying neurological processes respectively.

Re "reflex behaviour” vs "conscious behaviour" the latter requires higher neurological engagement. Conscious behaviour is dependent on self awareness and the contextual relationship between the individual and non-individual i.e. the rest of the world. It often requires the making of choices, though these might be environmentally conditioned. The label for reflex behaviour might be a bit unsubtle but it refers to observable behaviour that does not depend on consciousness or self awareness, with no higher order cognitive processes and no conscious choice about the response to stimulus. I think that the environment thing is a red herring as both are modified by environment and both by learning, although at radically different levels. The crunch is whether or not a fruit fly is conscious. Im just not going to disappear down that rabbit hole!

I throw these ideas into the mix.

P.

On 12 Feb 2020, at 09:42, David Osumi-Sutherland notifications@github.com wrote:

On 12 Feb 2020, at 09:06, Nico Matentzoglu notifications@github.com wrote:

Does this mean that animals without a developed consciousness like worms and flys exhibit only purely genetically programmed behaviors?

It seems that this will result in a big controversy.

Yep. Big can of worms. You’ve tangled consciousness (whatever that may mean) with the ability to learn, which even worms can do to some extent.

Can there be purely genetically programmed behavior? Isnt everything somewhat dependent on environmental factors (including social) but just to different degrees?

We already have this controversy in miniature inside NBO: reflex vs ‘conscious behaviour’ (terrible name for non-reflex in NBO). In a classical view - a reflex is (something like) an innate response to a particular environmental stimulus - i.e. there is an environmental component but the behaviour is not modified by experience. However, this is probably more a matter of degree rather than being categorical - even classical reflexes can be somewhat modified by experience.

So you would basically say:

GO:behaviors: behaviors that occur automatically under normal environmental conditions (availability of oxygen) based on some genetic mechanism (what if the behavior needs a predator to be triggered?) NBO:Behavior with a genetically programmed component NBO:Behavior with an environmentally programmed component NBO: courtship behavior subclass of NBO:Behavior with a genetically programmed component and NBO:Behavior with an environmentally programmed component Not sure I can make a lot of sense of drawing a clear line between genetically programmed vs not. But then again, I am not a biologist and defer to @dosumis https://github.com/dosumis here :P

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/obo-behavior/behavior-ontology/issues/101?email_source=notifications&email_token=AAA3RR3NCIPMFYKPHMPZ7X3RCO32PA5CNFSM4KSNMJWKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOELP7TNI#issuecomment-585103797, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAA3RR5YBWQ5JKR4JF7S643RCO32PANCNFSM4KSNMJWA.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

pmidford commented 4 years ago

I like the approach of splitting neurological process from observable process - but we shouldn't lose the distinction between observable processes and phenotypes as dispositions, which I thought was one of the most valuable contributions of NBO - an animal engages in particular processes but has a disposition to engage in them in particular ways, whether that disposition is due to genetic, disease state, or physical variation (e.g., missing a leg).  If the original intent of the two branches of NBO was to distinguish neurological processes from observable processes (rather than observable processes from genetically etc. determined phenotypes) than that was never clear to me and I think a step backward.  You can't reduce the observable process to a genetically determined neurological process because the environment always plays a role in the former.  Better to use the GO/NBO distinction to make that boundary than bring it inside NBO and conflate processes and dispositions.

How hard would it be to sell this to the GO folks?

-Peter Midford

On 2/12/20 03:43, PaulNSchofield wrote:

A lot of this is about behaviour as phenotype and behaviour as neurological process. So as a phenotype its observable, e.g. nurturing behaviour, psychotic behaviour, goal driven behaviour etc. You can observe it and you can measure it. The classes are therefore operationally defined.

Behavioural process is about neurological processes that result in observable behaviour. Doesn’t imply that all observable behaviour is genetic but that observable behaviour is dependent on the behavioural neurological processes. Maybe a specifically dependent occurrent? I hesitate to get into the philosophy here as there are those on this list better placed to make that call.

As David says, its about the mind/brain problem and thats why its extremely difficult to resolve. We talk about this problem in the NBO papers. Im not sure that Id defend my interpretation for the whole of NBO or GO ( Im sure we can find exceptions to this classification in both) , but I think that it would be interesting to view the domains of each as basically operationally defined phenotypes and underlying neurological processes respectively.

Re "reflex behaviour” vs "conscious behaviour" the latter requires higher neurological engagement. Conscious behaviour is dependent on self awareness and the contextual relationship between the individual and non-individual i.e. the rest of the world. It often requires the making of choices, though these might be environmentally conditioned. The label for reflex behaviour might be a bit unsubtle but it refers to observable behaviour that does not depend on consciousness or self awareness, with no higher order cognitive processes and no conscious choice about the response to stimulus. I think that the environment thing is a red herring as both are modified by environment and both by learning, although at radically different levels. The crunch is whether or not a fruit fly is conscious. Im just not going to disappear down that rabbit hole!

I throw these ideas into the mix.

P.

On 12 Feb 2020, at 09:42, David Osumi-Sutherland notifications@github.com wrote:

On 12 Feb 2020, at 09:06, Nico Matentzoglu notifications@github.com wrote:

Does this mean that animals without a developed consciousness like worms and flys exhibit only purely genetically programmed behaviors?

It seems that this will result in a big controversy.

Yep. Big can of worms. You’ve tangled consciousness (whatever that may mean) with the ability to learn, which even worms can do to some extent.

Can there be purely genetically programmed behavior? Isnt everything somewhat dependent on environmental factors (including social) but just to different degrees?

We already have this controversy in miniature inside NBO: reflex vs ‘conscious behaviour’ (terrible name for non-reflex in NBO). In a classical view - a reflex is (something like) an innate response to a particular environmental stimulus - i.e. there is an environmental component but the behaviour is not modified by experience. However, this is probably more a matter of degree rather than being categorical

  • even classical reflexes can be somewhat modified by experience.

So you would basically say:

GO:behaviors: behaviors that occur automatically under normal environmental conditions (availability of oxygen) based on some genetic mechanism (what if the behavior needs a predator to be triggered?) NBO:Behavior with a genetically programmed component NBO:Behavior with an environmentally programmed component NBO: courtship behavior subclass of NBO:Behavior with a genetically programmed component and NBO:Behavior with an environmentally programmed component Not sure I can make a lot of sense of drawing a clear line between genetically programmed vs not. But then again, I am not a biologist and defer to @dosumis https://github.com/dosumis here :P

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/obo-behavior/behavior-ontology/issues/101?email_source=notifications&email_token=AAA3RR3NCIPMFYKPHMPZ7X3RCO32PA5CNFSM4KSNMJWKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOELP7TNI#issuecomment-585103797, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAA3RR5YBWQ5JKR4JF7S643RCO32PANCNFSM4KSNMJWA.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/obo-behavior/behavior-ontology/issues/101?email_source=notifications&email_token=AAC7PDWWOQCQXUDDOFHC3YLRCPOGRA5CNFSM4KSNMJWKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOELQPDAQ#issuecomment-585167234, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAC7PDTVCKMYT4K6UQBPTN3RCPOGRANCNFSM4KSNMJWA.

cmungall commented 4 years ago

Note my proposal is not to split neurological processes from observable processes. It is to cast GO "behaviors" as something like circuits - the information flow through the organism. This would stand in some kind of 'underpins' relationship to organism-level behaviors. This relation would be complex and variable depending on the species.

dosumis commented 4 years ago

Focussing on only the top level class in NBO/GO right now:

I think the consensus here is that genetic vs environmental is not a useful distinction at this level (although I think we all want some restriction that tends to group thing with common molecular underpinnings). Correct?

Chris wrote:

Note my proposal is not to split neurological processes from observable processes. It is to cast GO "behaviors" as something like circuits - the information flow through the organism. This would stand in some kind of 'underpins' relationship to organism-level behaviors. This relation would be complex and variable depending on the species.

Is this too broad?

"The internally coordinated responses (actions or inactions) of animals (individuals or groups) to internal or external stimuli, via a mechanism that involves nervous system activity. " (= current def in GO)

By requiring a mechanism that involves the nervous system it (a) excludes non-animals (important because the molecular and cellular underpinnings are almost always unrelated (although see https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4640487/) (b) requires 'information flow' through 'circuits' be part of the process.

I think it is very naive to think you can separate out the circuit aspect. Behaviour is not disembodied. There are complex feedbacks between muscles, senses and circuits involved in almost any behaviour. Fun example - head stablization in birds: https://www.youtube.com/watch?v=aqgewVCC0k0 https://www.youtube.com/watch?v=GzT_k63T-rU

Think about the co-ordination of senses muscles and circuits involved. The hawk is an impressive example, but most or all movement behaviours involve these loops

There's lots of work on this in flies right now - how flies track targets in flight - identifying the wiring of feedback circuits involving muscles and sensory perception.

Peter wrote: If the original intent of the two branches of NBO was to distinguish neurological processes from observable processes....

Re - observable vs interpretation:
I agree we need to be careful about this at a more granular level (e.g. interpreting an observed movement as an escape response), but are you concerned that we'll occasionally include observed 'behaviours' that don't have any nervous system underpinning? Do you have any examples? Is this the flip side of thinking it is possible to separate out the nervous system circuit from the body?

On the phenotype side - we often have the problem that we don't know if the observed phenotype is due to an effect on the nervous system or on the musculo-skeletal system. If the behaviour term encompasses both then we don't have the problem.

@PaulNSchofield - the reflex vs conscious behaviour question should be moved out to another ticket as I don't think relevant to deciding upper-level def (short answer: we should have reflex, with some qualification, but the opposite is not conscious behaviour - otherwise we lose the animal behaviour community).

Summary:

matentzn commented 3 years ago

In the absence of a general solution and for practical reasons we assume now that the two root classes NBO:behavioral process and GO:behaviour refer to the same thing. NBO classes and GO classes continue to interleave for now. We need someone to come up with a proper behaviour modelling approach to deal with a clearer distinction here, also to sort the problem of delineating social activities like hiking and exercising from other ones.

DitchingIt commented 1 year ago

to sort the problem of delineating social activities like hiking and exercising from other ones

This discussion is now on COB and NBO is not attempting to incorporate such activities.

DitchingIt commented 10 months ago

This needs re-exploring in due course.