Design patterns for homeostasis terms

[dosumis]

(Issue moved from JIRA)

Homeostatis is not modelled properly in GO. Homeostatis != regulation. Here's my attempt to think it through from scratch - pretty much ignoring what we have in GO right now.

X homeostasis: a collection of negative feedback loops that serve to keep the level of X within some target range. The analogy of a central heating system with a thermostat is sometimes used to illustrate this.

All instances of homeostasis have a target range or level. (In our heating system analogy, set by the thermostat). We could call this the homeostatic range or homeostatic level, which we could model (ignoring BFO strictures) as a quality inhering in homeostasis.

Each feedback loop includes some means of sensing the level of X linked to some means of changing the level of X back towards the target range.

X homeostasis <- part_of regulation of X levels involved in X homeostasis <- part_of detection of X levels involved in X homeostasis

In some cases, endogenous mechanisms exist that change the target range of X. Endogenous pyrogens raise body temperature. Increasing adrenalin raises blood pressure, blood sugar and blood fatty acid levels. In our analogy, some process turns the thermostat up. This is something distinct from homeostasis itself, and some of the processes involved (transduction of adrenaline signal?) are clearly not part of homeostasis. It seems odd to call this regulation of homeostasis. It would be more natural to see it as regulation homeostatic range or level. I suggest:

regulation of homeostatic X level: Any process that regulates the target range or level towards which an instance of X homeostasis tends.

== Parentage:==

Should X homeostasis go under regulation of X? (it certainly results in some regulation of X levels), or be out on its own under BP. I'm agnostic.

== Caveats: ==

I think that the story I've told here works pretty well for a bunch of classical, organism level physiological processes - many of which are surprisingly poorly represented in GO. We should however, flesh out some examples with experts and see if it meshes with how they see the biology. There must surely be a paper in this.

This story may not apply as effectively to all other uses of the term homeostasis - e.g. when applied to cell number. I also suspect that regulation of some intracellular chemical levels is a more complicated than this story. Ca2+, for example, has complex intracellular dynamics that play a critical role in signalling. Metabolite levels can change dramatically in exercising muscle. Should the term homeostasis be used in these cases.

AIs:

1. Draft design patterns and apply to new set of terms for blood sugar homeostasis.
2. Get expert input from medics/physiologists/system modelling people.
3. Once agreed, alert annotators that existing annotations to sugar homeostasis should be reviewed and moved.

[dosumis]

Comment from @cmungall, pasted from email:

We should however, flesh out some examples with experts and see if it meshes with how they see the biology. There must surely be a paper in this.

Definitely. But ideally this would be more than a "we improved some textual definitions in the GO" type paper, or a "here is a formalism that doesn't really have any bearing on what we're doing" paper, and we could work out some use cases where improved modeling makes a big difference… but this will take more time.

Maybe talk to Bernard de Bono?

If we want to get this right should we loop in some systems bio / dynamic modeling people? Ask Nicolas Le Novere?

Also, how does this fit with disease? Sometimes diseases are conceived of as deviations from homeostasis. This is maybe sometimes an attempt to lend a formal mechanistic engineering legitimacy onto something that's more fuzzy, but I think there is value in extending our thinking to non-canonical processes here. I think you're doing this already (your blood glucose example), just wanted to make this more explicit.

[dosumis]

Discussion with @rlovering - from email:

Hi Ruth, On 30 Apr 2015, at 09:17, "Lovering, Ruth" wrote:

Hi David Thanks for including me in this discussion. It does sound like someone from UCL might be helpful. Getting the ontology right for one homeostatic process might provide the example case required. Thinking along those lines please note that in 2007 (I think) there was a ontology workshop which developed the blood pressure ontology domain, and there are 137 child term for regulation of blood pressure. Here there is no mention of a homeostatic process, even though obviously the body does try to maintain blood pressure within a specific range. Wrt your comment: X homeostasis: a collection of negative feedback loops that serve to keep the level of X within some target range. I think this should be a collection of negative and positive feedback loops?

Wouldn't positive feedback loops tend to act against homeostatsis? I think it makes sense to broadly refer to the mechanisms as negative feedback, even if the details may involve more complex dynamics and thresholds than simple negative feeback loops.

Talking to David yesterday, there is also the decision to raise or lower from what the body knows is the optimal level in order to meet the need of the situation. Which is perhaps why the blood pressure ontology hasn’t got a homeostasis level. But is this always a different mechanism? Potentially the detecting/stimulating/inhibiting mechanisms are regulating the activity/process which usually keeps the ‘levels’ set within the ’normal’ range. I think the descriptive terms in the blood pressure domain might be worth considering to see how these are formatted.

Definitely worth looking. But the ' decision to raise or lower from what the body knows is the optimal level in order to meet the need of the situation' can involve hormones that are outside of the normal regulatory mechanisms - e.g. adrenaline raises blood pressure.

We could discuss the lipid level regulation mechanism in the CVG group, but obviously in most people the regulation of this is overwhelmed! But I think there is an ideal range that we are able to keep our blood lipid levels within.

Definitely relevant. This discussion prompted by Rachael's annotation of this paper.

Perhaps blood glucose homeostasis would be a good example to start with. The biology is well known and, as for blood pressure and blood lipid levels, the medical significance is huge. But GO doesn't represent it in any useful way right now.

[dosumis]

Made a start:

https://github.com/geneontology/design_patterns/blob/master/draft_patterns/chemical_homeostasis_in.yaml

[hdietze]

For reference, here is the TermGenie pattern js file for chemical homeostasis and cellular chemical homeostasis: https://github.com/geneontology/termgenie/blob/master/TermGenie/TermGenieRulesDynamic/src/main/resources/rules/go/chebi/chemical_homeostasis.js

[dosumis]

The homeostatis terms themselves are less problematic that the 'regulation of homeostatis terms'. Hard to see what regulation of homeostatis can be apart from regulation of homeostatic level.

Inferred classes under these are really just regulation of levels :

[dosumis]

For glucose homeostatis example:

Reviewing direact annotation with 'glucose homeostatis' - first 5 papers selected all refer to regulation of blood glucose levels by insulin.