Open Issue : Pathway Step

Definition:
A step in a pathway.

Comment:
Multiple interactions may occur in a pathway step, each should be listed in the STEP-INTERACTIONS property. Order relationships between pathway steps may be established with the NEXT-STEP slot. This order may not be temporally meaningful for specific steps, such as for a pathway loop or a reversible reaction, but represents a directed graph of step relationships that can be useful for describing the overall flow of a pathway, as may be useful in a pathway diagram.

Example:
A metabolic pathway may contain a pathway step composed of one biochemical reaction (BR1) and one catalysis (CAT1) instance, where CAT1 describes the catalysis of BR1.

Current Issues :

 * 1) Pathway Step implies a direction ( due to next- step ). So when there is a reversible reaction in the pathway step, its direction needs to be specified. This is already a requirement for KEGG. How about others?
 * 2) Pathway Step is far complicated beyond the requirements of any data provider.As it is  a pathway can have a mixed bag of interactions,(sub)pathways and pathway steps. Pathway steps then can contain its own set of interactions and pathways. Pathway steps basically implement a full nested graph. This structure is extremely difficult to handle and navigate and unnecessarily so, as no one uses such an extensive pathway step representation.
 * 3) An interaction within a pathway-step which in turn contained by a pathways is still a component of that pathway. This is not currently reflected in OWL.
 * 4) Semantics of a pathway step containing pathway is not clear. Also I am not aware of any data provider requiring this.
 * 5) INOH requires ability to annotate pathway steps with evidence. In this sense the evidence is in the form A is followed by B, so it truly belongs to the pathway step.

Proposed Solution(s):
To agitate some discussion, I propose a drastic solution of limiting pathway steps to have one conversion and multiple controls. Let's start from this point and try to relax the solution based on the requirements of the data providers. Having always one conversion makes it possible to address all the issue 1, with a simple property on the step indicating left to right or vice versa. As with the 2 and 4 simplified structure makes those issues null.Issue 3 can be fixed my Declaring a transitive parent to properties PATHWAY-COMPONENTS and STEP-INTERACTIONS Issue 5 can also be easily fixed with allowing an evidence pathway step. What do you think?

--

Feedback from Suzanne Paley:
Suzanne : Is directionality for a step required? It seems to me that it is only necessary if a step does in fact have a NEXT-STEP value, or is a NEXT-STEP for some other step. We happen to have a few pathways which consist of several reversible reactions which interconvert the same few species, and can be strung together in any order -- we don't supply any ordering or direction information for the reactions in these pathways. Would this still be allowed? Emek: There is nothing that would prevent you from putting these into a pathway without pathway steps.

Suzanne : How would superpathways be handled?

Emek: Pathways can still have pathways as their components. However you can not put sub-pathways into pathway steps. You can still explicitly link sub-pathways, by linking specific interactions but can not vaguely say this sub-pathway is followed by this pathway.

Suzanne : So if I wanted to say that some protein modification step led to some standard pathway for ubiquitinylation and degradation of the protein, how might I express that? (note that we don't actually have any pathways like this right now, but it's the kind of thing we are thinking we will want to be able to handle in the future) Emek: Very nice point. It is great to have such feedback. Addressing degradation and generics/templates is a related problem that is still open in biopax. However for ubiquitinylation one could define A->[1]->A(p)->[2]->A(p,ub) and then put [2]into ubiquitinylation pathway and then link [1] and [2] with pathway steps. Now one might say that I cheated and in fact I did. But using pathway steps for this (linking a phosphorylation to a vague definition of 'protein degradation') can also be problematic and quite ambiguous. What we really need for this type of data is a way to formally represent generic interactions, an issue already raised by Ken Fukuda, and I will post a separate e-mail for this.

---

Feedback from Peter Karp:
In response to : " Pathway Step implies a direction ( due to next- step ). "

Peter : It implies a direction, but ONLY WITHIN THIS PATHWAY. The reaction could be used in other directions in other pathways or independent of pathways.

Emek : I agree. As pathway steps are not used across pathways this implicitly holds.

--

Feedback from Dan Corwin:
A pathway is an abstract (informational) object, not unlike a story. The steps in it are also abstract objects, not unlike chapters or paragraphs. Their order, if any, has to do with the sequence of sub-presentations used by the pathway's author. The drastic solution would indeed limit what each such sub-presentation can cite, but I don't see what benefits that would offer, or to whom, or the problem being "solved".

NEXT-STEP has nothing whatsoever to do with the directionality (if any) of the topics it cites - which could be interactions, or other pathways, or whatever else BioPAX chooses to allow within its presentation rules for pathways. But we should all acknowledge at this level we are talking about a style guide for a totally artifical hierarchy of invented models, not the details of real physical entities to which such models might refer.