Naming discussion: Pathway vs Network

[William Hayes]

Hi all,

We've been struggling with naming to distinguish the database of network edges as represented by BEL Statements and curated networks created from this database of edges for a specific biological question.

BKN/EdgeStore:  BEL Nanopub BEL Statements and the expanded/transformed edges that can be created from BEL Statements (e.g. complex(p(HGNC:F3),p(HGNC:F7)) is expanded to complex(p(HGNC:F3),p(HGNC:F7)) hasComponent p(HGNC:F3)).

Example Pathway/Network:  http://www.causalbionet.com/Home/NetworkVisualization/#/!/networkGraph?ID=angiogenesis_2.0_hs

I haven't found a good definition of biological network vs pathway. 

I am proposing that we use Pathway for a network like the Angiogenesis network represented in the example from CausalBioNet (CBN) and Network for the BKN (Biological Knowledge Network) representing the full set of edges/relationships in the biology space (of which we may only have a subset based on current knowledge).  

This would lead to renaming the Network Store (proposed component of OpenBEL) to Pathway Store, and referring to those types of networks (like Angiogenesis from CBN) as Pathways instead of networks.  We would then have semantic distinction between a database of edges (BKN) as a network and a curated set of edges for a specific biological context as a Pathway.

Biological Network:  known set of edges representing biological interactions and relationships

Pathway:  curated set of edges for a specific biological context (species, tissues, disease, development stage, etc)

[Dexter Pratt]

Hi William,

I’ve been thinking about exactly this issue recently, and I basically agree except that I would make definition for an ideal “pathway” network tighter.

Ideally, a “pathway” is a model in which all of the paths within the model represent valid sub-models intended by the authors. Even better, causal effect of some or all of the paths is supported by empirical evidence.

This corresponds to a diagram that an expert in a given domain would create, not a team of curators.

If the model contains the path BRAF->MAP2K1->MAPK3, the expert believes this causal sequence is experimentally supported within the context of the model. And better, they have supported the model with an included or referenced empirical relationship BRAF->MAPK3.

Many networks that are called “pathways” contain edge that do meet this criterion, even though the network is focused on a specific mechanism or context and all of the edges are supported by evidence.

Here is the critical distinction: if I am looking for paths to explain or predict effects between entities A and D, I will be willing to consider a longer path within a model meeting my strict definition.   My trust in paths of comparable length that no expert has ever explicitly vouched for is much lower, even if the component edges are well supported.

I believe that adding this path-level literature support to our analyses is critical, especially in communicating with the wet-lab experimentalists who need focused hypotheses that they can take to the lab.

If the question is some variation of “how are the observed changes related to the perturbation in this experiment?” the biologist needs a nuanced answer:

Here are paths that are:

1. explanations of some (or all) of the observed changes in strict pathway models from experts: the literature provides a fully supported explanation.

2. are minor variations on known strict paths - somewhat novel, potentially high-priority hypotheses to test

3. are pieced together from well supported edges (but short)

4. are yet more speculative… (but even shorter)

More later.  My team is actively working on implementing this type of query for deployment as part of the NDEx interface.

Dexter

--
You received this message because you are subscribed to the Google Groups "openbel-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openbel-discu...@googlegroups.com.
To post to this group, send email to openbel...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openbel-discuss/895f7af5-b825-4118-b3d6-aed8fd0a55a1%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[William Hayes]

We really need a much richer vocabulary for biological network analysis.  Your definition of pathway is also needed, but it's different from the focused biological context, curated subnetwork that I'm trying to refer to.  

[Dexter Pratt]

Yes, I think there is an important ontology networks to be defined. I’m also interested in the class that you reference - for context X, a given path through that type of network that is also of context X is more credible than a path from a network in another context or from a large aggregate network.

On Feb 27, 2017, at 8:56 AM, William Hayes <william...@gmail.com> wrote:

We really need a much richer vocabulary for biological network analysis.  Your definition of pathway is also needed, but it's different from the focused biological context, curated subnetwork that I'm trying to refer to.  

--
You received this message because you are subscribed to the Google Groups "openbel-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openbel-discu...@googlegroups.com.
To post to this group, send email to openbel...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/openbel-discuss/d3cb29ab-1a5b-4f49-abec-71ba5b242a26%40googlegroups.com.

[Anselmo Di Fabio]

How about this; let's keep pathway simple: A "pathway" is an ordered series of actions that lead to an expected outcome.  Now, a network can be more complex: 

A network is a collection of outcomes (pathways) gathered together to define greater relationships. These relationships do not need to reflect connected components.

[Dexter Pratt]

I think that is a good concept for the ontology, but I don’t think it should be called “pathway" because that word has a very different and unfortunately ambiguous meaning to biologists.

I think “path” should be the graph theory definition, see https://en.wikipedia.org/wiki/Path_(graph_theory)

The “pathway” that you propose would be a path with causal, directed edges, and I would suggest “causal path” as the name of that term.

I think that biologists use “pathway” in an informal way to indicate networks that are like the diagrams that are drawn in review papers, etc. Wikipathways, for example.

That concept is so ambiguous that I now suggest we not use the word “pathway” as a formal term - no matter what we do, it will be confusing.

So I need a different name for my “strict” pathway term, defined as "a qualitative causal model network in which the author asserts that all causal paths are valid sub-models."

(*sigh*  - “binds to” edges and other non-causal edges suggest that I might want to loosen that definition a bit to get a more useful term)

In general, I would avoid stepping on existing graph theory / network terminology.  And BEL graphs or derived models are one case of qualitative causal model, so we should be careful to distinguish (1) terms that apply to any qualitative network model and (2) terms that have a specific BEL meaning.  

Dexter

On Feb 28, 2017, at 2:53 PM, Anselmo Di Fabio <adif...@adsworks.com> wrote:

How about this; let's keep pathway simple: A "pathway" is an ordered series of actions that lead to an expected outcome.  Now, a network can be more complex: 

A network is a collection of outcomes (pathways) gathered together to define greater relationships. These relationships do not need to reflect connected components.

--
You received this message because you are subscribed to the Google Groups "openbel-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openbel-discu...@googlegroups.com.
To post to this group, send email to openbel...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/openbel-discuss/53a4f96c-48c1-4c6a-b21a-efe071e7c920%40googlegroups.com.

[Anselmo Di Fabio]

Hmmm. I see what you are saying... Can we use the term "outcome" instead of pathway?  At the end of the day are we not looking for causal outcomes?  I used it in the definition of network.  I think we are getting closer.

You received this message because you are subscribed to a topic in the Google Groups "openbel-discuss" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/openbel-discuss/epOmF0-OmIE/unsubscribe.
To unsubscribe from this group and all its topics, send an email to openbel-discu...@googlegroups.com.

To post to this group, send email to openbel...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/openbel-discuss/0B80A5D4-95B7-4269-9633-A97F86B69BFC%40gmail.com.

For more options, visit https://groups.google.com/d/optout.

--

Anselmo Di Fabio || Man, Myth, CTO 
Email: adif...@adsworks.com
Phone: +1 732.764.8844 

 
MAKE IT RIGHT 

www.adsworks.com || LinkedIn 

The information transmitted is intended only for the person or entity to which it is addressed and may contain confidential and/or privileged material. Any review, retransmission, dissemination or other use of, or taking of any action in reliance upon, this information by persons or entities other than the intended recipient is prohibited. If you received this in error, please contact the sender and delete the material from any computer.

[Dexter Pratt]

Would “outcome” be the terminal node of a causal path? Multiple causal paths can share a terminal node, so one could say they have the same “outcome”?

I like words like “outcome” because they are different than general graph vocabulary. “outcome” has a connotation that feels like it is specific to qualtative causal models.

From my NDEx point of view, I need to deal with a heterogeneous set of networks, some of which are qualitative causal networks and of those, some have BEL semantics. So I’m advocating for a policy which would avoid different meanings for terms depending on which network type was being considered. So all networks have a common meaning of node, edge, path, network, node attribute, edge attribute, etc. I think we should extend that ontology in appropriate layers of increasing specificity.

To view this discussion on the web visit https://groups.google.com/d/msgid/openbel-discuss/CANsEe%2BrFVL3wxDn49LmdJgkfgjLF7iTLRq8NkM%2BummU5EfXK8A%40mail.gmail.com.

[Anselmo Di Fabio]

Okay, after sleeping on it a bit I feel this may be a good take: Since there is reservation on using the word pathways I am thinking we move to the work "Trails" or "Causal Trails" and each Trail can then terminate in an "Outcome"  The term allows for the condition of multiple Trails leading to the same outcome.  Also, it allows for some separation from the more general term of pathways.