Closed RichardBruskiewich closed 2 years ago
Great start, but we need to flesh this out more. More example edges.
In particular, I want to know where these MOA nodes are coming from?
The hierarchy here is limited, this seems more like something modeled as a qualifier on an edge or node: https://www.ebi.ac.uk/ols/ontologies/ncit/terms?iri=http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FNCIT_C54680
Would you consider the CHEBI 'role' hierarchy in scope here?
E.g here is your Paclitaxel example in CHEBI
We have a project with @choyt called CHIRO that axiomatizes the roles, essentially providing outward links to other nodes, allowing you to shortcut
E..g
then using a property chain we can eliminate the role middleperson
I would opt for this direct form of modeling, where we connect directly from the drug to the node of interest, and no role/MOA middleperson
I am flattered by the mention, but I am not the choyt you are looking for.
Thanks @cmungall for your feedback about about removing MOA as a middle node. In my latest commit https://github.com/biolink/biolink-model/pull/627/commits/d24f79046bad0950b85dc3191d8f4c9730cb0b18 to PR #627, I've converted "MechanismOfAction
" into the predicate mechanism_of_action
which would be used in a simpler DrugToBiologicalEntityAssociation
which captures the semantics of the situation more directly.
Yes indeed I meant to tag Charles hoyt.. but that is awesome advice anyway! We are trying to find the right balance between overloading subclass yet retaing its benefits. In fact we were just discussing the need for composition pattern for modeling drugs and treatments in our hackathon today!!
On Tue, Feb 2, 2021, 17:06 Curt Hoyt notifications@github.com wrote:
Also, I recommend using a strongly-typed hierarchy rather than a class-based polymorphic approach. Composition is more generally extensible than inheritance, and mixins often end up multiplying layers of abstraction without a concomitant increase in clarity.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/biolink/biolink-model/issues/626#issuecomment-772128836, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAAMMOOZFEYLIPRVZHGNGR3S5COSPANCNFSM4W7ROS5A .
Is DrugToBiologicalEntityAssociation appropriate association? Mechanism of action is also applicable to chemical substances (biochemical probes that are not drugs) and possibly proteins that may act as natural ligands. Maybe MolecularEntityToBiologicalEntityAssociation?
Here's an example of a very specific MoA https://www.accessdata.fda.gov/drugsatfda_docs/label/2003/021588lbl.pdf (top of page 2). How would you model something like this?
I think there is some reason to distinguish this from the previous discussion of Mechanism of Action https://github.com/biolink/biolink-model/issues/570. While I think the intent in that issue was to model pharmacology as predicates within Biolink descending from affects
, here the drug to (typically) protein target relationship connotes additionally that this target engagement is principally responsible for the therapeutic effect of the drug. So while imatinib inhibits many protein kinases, its inhibition of BCR-ABL is why it is a useful AML therapeutic. As a further case to consider, both loperamide (Imodium) and morphine act as opioid-receptor agonists acting on the μ-opioid receptors, and while both induce constipation, only the latter affords pain relief --- so I am not sure if further modeling of mechanism of action including context like tissue of action is required to properly annotate these two MoAs.
What is the status of this?
@suihuang-ISB commented on "mechanism of action" in this ticket: https://github.com/NCATSTranslator/testing/issues/93#issuecomment-896364310
We have discussed this issue at the chem-info WG call and have agreed that mechanism of action should be modelled as a path in a knowledge graph or even as a subgraph in more complex MoA cases. Rather then to have a single node to represent MoA, we propose to have a dedicated attribute, biolink:mechanism_of_action
, that would indicate which nodes and edges represent given mechanism of action.
node attribute or edge attribute? I assume the latter since the MOA will be contextual?
On Wed, Sep 15, 2021 at 8:41 AM vdancik @.***> wrote:
We have discussed this issue at the chem-info WG call and have agreed that mechanism of action should be modelled as a path in a knowledge graph or even as a subgraph in more complex MoA cases. Rather then to have a single node to represent MoA, we propose to have a dedicated attribute, biolink:mechanism_of_action, that would indicate which nodes and edges represent given mechanism of action.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/biolink/biolink-model/issues/626#issuecomment-920134830, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAAMMOMM4POHR4E4LFWFYLTUCC5D7ANCNFSM4W7ROS5A . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.
Amateur here. @suihuang-ISB made the point that MoAs are often controversial, which could be modeled by a Range of theories. However, I would propose that any piece of knowledge considered controversial by reasonable, knowledgeable people should simply be left out of the graph, rather than being reflected in its schema. How is this currently handled?
Problem
No category that adequately captures the spirit of NCIT:C54680, Mechanism of Action: The mechanism by which a pharmacologically active substance produces an effect on a living organism or in a biochemical system.
Describe the solution you'd like
Add a "Mechanism of Action" category plus some supporting biolink:Association types and possibly, predicates, allowing 1) placeholder of Drug -...-> Mechanism of Action -...-> subgraph of descriptive edges (e.g. anatomical site of action, process, molecular targets, etc.)
Working group (or team) this request originate from:
Requested by Trung of LinkBrokers during February 2021 Relay.
Additional context
https://bioportal.bioontology.org/ontologies/ATC/ - "...classification of active ingredients of drugs according to the organ or system on which they act and their therapeutic, pharmacological and chemical properties...."
people would recognize “antineoplastic” as the MoA (not very specific but it’s a general category (note: Antineoplastic: Acting to prevent, inhibit or halt the development of a neoplasm (a tumor). An agent with antineoplastic properties.)
several CHEMBL.MECHANISM mapped terms in the Biolink Model but mostly under specific molecular-level categories, i.e. a predicate like biolink:physically_interacts_with
"mechanism of action" is a general concept node type, from which one would hang off several qualifying Association edges (e.g. anatomic site of action, process target, chemical reaction, etc.)
specific to a cell type or cellular process or cell surface receptor, e.g.
Tagged members for discussion
@vdancik @cbizon @mbrush @cmungall