#16 Relations between entities from Molecular Interactions ontol

Larry Hunter

The Human Proteome Organization Standards Initiative has a proposed a community guideline regarding how to fully describe a molecular interaction experiment, called MIMIx, and an interchange standard for describing such experiments, PSI-MI (currently in version 2.5.3). This work was originally described in Hermjakob, et al. (2004) The HUPO PSI's molecular interaction format—a community standard for the representation of protein interaction data Nat. Biotechnol, . 22, 177–183 and the 2.5 release is described in Kerrien et al., Broadening the horizon-level 2.5 of the HUPO-PSI format for molecular interactions, BMC Biol. 5 (2007), p. 44. This standard is used by a variety of important public resources, including MINT, IntAct, DOMINO, CORUM and as the basis for an aspect of the BioCreative II text mining competition.

Recently, the Structured Digital Abstract initiative at FEBS letters (see Ceol, et al., Linking entries in protein interaction database to structured text: The FEBS Letters experiment, FEBS Letters Volume 582, Issue 8, 9 April 2008, Pages 1171-1177 adopted the PSI-MI as its source of relationships between proteins, as well as for describing the evidence for each assertion.

In the interests of ensuring compatibility between the RO and these important community resources, the relationships from the PSI-MI ontology need to be assessed for inclusion in the RO.

There are relevant portions of the PSI-MI are the molecular interactions (MI:0190 and children) and possibly the interaction detection method (MI:0001 and children), participant identification method (MI:0002 and children) and feature detection method (MI:0003) as kinds of supports relations.


  • Chris Mungall
    Chris Mungall

    Logged In: YES
    Originator: NO


    This is the section you speak of:

    ./ MI:0000 molecular interaction
    ..po MI:0190 interaction type
    ...is_a MI:0208 genetic interaction
    ....is_a MI:0794 synthetic interaction
    ....is_a MI:0795 asynthetic interaction
    ....is_a MI:0796 suppressive interaction
    ....is_a MI:0797 epistatic interaction
    ....is_a MI:0798 conditional interaction
    .....is_a MI:0802 enhancement interaction
    ....is_a MI:0799 additive interaction
    .....is_a MI:0802 enhancement interaction
    ....is_a MI:0800 single nonmonotonic interaction
    ....is_a MI:0801 double nonmonotonic interaction
    ...is_a MI:0218 physical interaction
    ....is_a MI:0407 direct interaction
    .....is_a MI:0195 covalent binding
    ......is_a MI:0408 disulfide bond
    ......is_a MI:0556 transglutamination reaction
    .....is_a MI:0414 enzymatic reaction
    ......is_a MI:0192 acetylation reaction
    ......is_a MI:0193 amidation reaction
    ......is_a MI:0194 cleavage reaction
    .......is_a MI:0212 lipid cleavage
    ........is_a MI:0198 defarnesylation reaction
    ........is_a MI:0200 degeranylation reaction
    ........is_a MI:0201 demyristoylation reaction
    ........is_a MI:0202 depalmitoylation reaction
    .......is_a MI:0570 protein cleavage
    .......is_a MI:0571 mrna cleavage
    .......is_a MI:0572 dna cleavage
    ......is_a MI:0197 deacetylation reaction
    ......is_a MI:0199 deformylation reaction
    ......is_a MI:0203 dephosphorylation reaction
    ......is_a MI:0204 deubiquitination reaction
    ......is_a MI:0207 formylation reaction
    ......is_a MI:0210 hydroxylation reaction
    ......is_a MI:0211 lipid addition
    .......is_a MI:0206 farnesylation reaction
    .......is_a MI:0209 geranylgeranylation reaction
    .......is_a MI:0214 myristoylation reaction
    .......is_a MI:0216 palmitoylation reaction
    ......is_a MI:0213 methylation reaction
    ......is_a MI:0217 phosphorylation reaction
    ......is_a MI:0220 ubiquitination reaction
    ......is_a MI:0557 adp ribosylation reaction
    ......is_a MI:0558 deglycosylation reaction
    ......is_a MI:0559 glycosylation reaction
    ......is_a MI:0566 sumoylation reaction
    ......is_a MI:0567 neddylation reaction
    ......is_a MI:0568 desumoylation reaction
    ......is_a MI:0569 deneddylation reaction
    ......is_a MI:0701 dna strand elongation
    ......is_a MI:0844 phosphotransfer reaction
    ......is_a MI:0871 demethylation reaction
    ......is_a MI:0881 nucleoside triphosphatase reaction
    .......is_a MI:0882 atpase reaction
    .......is_a MI:0883 gtpase reaction
    ...is_a MI:0403 colocalization

    These are not relations

  • Larry Hunter
    Larry Hunter

    Logged In: YES
    Originator: YES

    I know what you mean, but take a look at the way they are used in the resources I cited above. For example, the FEBS SDA for doi:10.1016/j.febslet.2007.12.036 is
    "Aurora B (uniprotkb:Q96GD4) phosphorylates (MI:0217) MgcRacGAP (uniprotkb:Q9H0H5) by protein kinase assay (MI:0424)" and the related record in MINT
    http://mint.bio.uniroma2.it/mint/search/interaction.do?interactionAc=MINT-6166761 also treats these types of physical interactions as relationships.

    We will obviously have to make some kind of mapping from the PSM-MI interaction classes to well-defined relations, but it seems to me that this mapping is fairly straightforward. The terms of the form "X interaction" map to a relation something like "has-a-X-interaction-with" The subclasses of "direct interaction" that are of the form "<x>tion reaction" map to a relation something like "<x>ates: (e.g "adp ribosylation reaction" -> "adp ribosylates"), which is what FEBS did with "phosphorylates". The few subclasses not of that form (e.g. "cleavage reaction") have natural formulations as relations as well (in this case, "cleaves").

    Perhaps the most complex issue is quantification. Is "Aurora B phosphorylates MgcRAcGAP" all-some? Probably not. Personally, I am comfortable with making instance relations from these, and letting the ontologists figure out when class level relations are justified. B

    I think the widespread use of these not-quite-appropriate not-quite-relations in this way really cries out for the RO to propose a better alternative.

  • Chris Mungall
    Chris Mungall

    Logged In: YES
    Originator: NO

    neat viewer.

    I don't think there's anything new here that we don't have to deal with the the BP xps

    e.g. phosphorylation is a process. we need something stronger than has_participant in order to specify the role each participant plays. We don't know quite the best way to do this that is a balance between a relation hierarchy that is redundant with the process hierarchy. The existing bp_xp_cell definitions are our best guess so far

  • Logged In: YES
    Originator: NO

    Can these be represented, schematically, as

    phosphorylation-of-A-by-B subClassOf phosphorylation
    has_participant some (A and has_function K-Acceptor)
    has_participant some (B and has_function kinase_function)
    has_participant some (ATP and has_function K-donor)
    has_participant some (ASerinePhosAt22 has_role PTM_A)

    We are missing that A is present at the beginning and ASerinePhosAt22 and vice versa at the end,
    and that the ASerinePhosAt22 is derived_from the A.

    The first two might be captured with time sensitive participation relations along the lines of p has_beginning_participant c =def exists t | c participates in p at t1 , t1 < t, and c not participates in p at t1, t1 > t.

    I'm not sure about the third bit that the ASerinePhosAt22 is derived_from the A. Works fine in FOL, I'm thinking of how to do it in OWL.

    No reason not to have B phosphorylates A as a macro that expands in to the above, and can be queried out of the above.