Requesting new experimental role terms and one interactor type for genetic interactions

[chris-grove]

The Alliance of Genome Resources Interactions working group would like to request several new terms for genetic interaction interactors.

We realize that there already exist the terms:

'suppressor gene' 'suppressed gene'

under the 'experimental role' branch. As it is more accurate to refer to the perturbation (genetic, chemical, environmental) as the 'suppressor' or 'suppressed' entity we are proposing to create the following terms as children of the 'experimental role' term:

'suppressor perturbation' def: 'A perturbation that suppresses the phenotype of another perturbation.'

'suppressed perturbation' def: 'A perturbation whose phenotype is suppressed by another perturbation.'

'enhancer perturbation' def: 'A perturbation that enhances the phenotype of another perturbation.'

'enhanced perturbation' def: 'A perturbation whose phenotype is enhanced by another perturbation.'

'epistatic perturbation' def: 'A perturbation whose phenotype masks the phenotype of another perturbation.'

'hypostatic perturbation' def: 'A perturbation whose phenotype is masked by the phenotype of another perturbation.'

and we propose an additional term as a child of 'interactor type':

'perturbation' def: 'A modification of an organism and/or its environment that may be capable of inducing or modifying a phenotype in that organism. This may include genetic perturbations, such as a gene mutation or RNA interference, chemical perturbations, such as the introduction of a drug or chemical, or environmental perturbations, such as a change in ambient temperature or humidity.'

As 'perturbation' is such a generic term, we may want to instead label the term 'organism perturbation' or something to that effect (both in the term label and in the experimental role term definitions).

Feedback is welcome.

[pporrasebi]

The concept of entering a perturbation as an interactor I think enters in direct conflict with the concept of interaction representation we have taken at PSI-MI so far. We aim to capture different types of interctions -be it genetic, physical or causal- among physical molecules or genes. Perturbations that either affect or are derived from these interactions are captured in different ways: as features of interactors/participants (see the 'mutation' branch under MI:0118, https://www.ebi.ac.uk/ols/ontologies/mi/terms?iri=http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FMI_0118), as agonist/antagonist annotations (MI:0625/MI:0626), using the biological role to depict an effector/target relationship type or extending this via causal statements or mechanisms (see the whole branch devoted to this here:https://www.ebi.ac.uk/ols/ontologies/mi/terms?iri=http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FMI_2233&viewMode=All&siblings=false).

I believe having perturbation as an interactor, especially the way you have defined it, enters in direct conflict with most of the curation practice we have been following until now, so I would rather not do it. I am happy to have a conversation about how we can best accommodate for the use cases you have in mind, to explore how they can be rerpesented using current terms and practice, if possible.

[chris-grove]

@pporrasebi OK, thanks for your feedback.

Ultimately, all genetic interactions are phenotype-based interactions between genetic perturbations of genes (I'm happy to discuss in detail and would be very interested in learning of examples that don't fit this criteria). I suppose we can stick to using "gene" as the interactor for now, but it isn't quite accurate. Our working proposal for genetic interactions is to submit the primary interactors as genes (with type "gene") and also submit the identifiers for individual genetic perturbations for each gene (for example using WormBase ?Variation/Allele class object identifiers like "WBVar00012345") in columns 26 and 27 of the MITAB file ("Annotations for interactor A/B" columns) in agreement with prior discussions with @sandraorchard. Including these identifiers will be critical for accurate capturing/reporting/interpretation of genetic interactions from WormBase and FlyBase and maybe others in the future.

The most accurate way to express the nature of interactors (genetic perturbations) in a genetic interaction is to say that a particular perturbation (as interactor) has the role of being the suppressor/suppressed, enhancer/enhanced, etc. We can kludge it for now and just refer to an interactor role of "suppressor interactor" and "suppressed interactor", for example, but actually those descriptors are most accurately applied to the genetic perturbations themselves and not, say, the genes (hence why I'm somewhat against referring to "suppressor gene" or "suppressed gene").

Also, as you will see in the upcoming proposal for changes to the genetic interaction branch of the ontology/CV, we are trying to develop it with other types of phenotype-based interactions in mind, like drug-drug interactions, chemical-genetic interactions, environmental-genetic interactions, etc. So the type and nature of interactors will need to expand beyond genes to other types of entities like chemicals/small molecules, environments (or environmental conditions), and so on, all of which would fit the definition of "perturbation" that could be applied universally to these types of interactions.

Maybe we can simply make replace the word "perturbation" with "interactor" for the time being?

'suppressor interactor' def: 'A interactor that suppresses the phenotype of another interactor.'

'suppressed interactor' def: 'A interactor whose phenotype is suppressed by another interactor.'

'enhancer interactor' def: 'A interactor that enhances the phenotype of another interactor.'

'enhanced interactor' def: 'A interactor whose phenotype is enhanced by another interactor.'

'epistatic interactor' def: 'A interactor whose phenotype masks the phenotype of another interactor.'

'hypostatic interactor' def: 'A interactor whose phenotype is masked by the phenotype of another interactor.'

[chris-grove]

@pporrasebi Any updates on this? We will be generating a PSI-MITAB file for genetic interactions soon for the Alliance of Genome Resources and we will need the interactor types listed above:

suppressor interactor suppressed interactor enhancer interactor enhanced interactor epistatic interactor hypostatic interactor

Do you think we could get those terms?

[pporrasebi]

Sure, I was waiting for your feedback on the proposed definitions. If you are fine with them I can create the terms as soon as I jhave some time within the next few days.

[chris-grove]

@pporrasebi We can go ahead with the definitions proposed above:

'suppressor interactor' def: 'A interactor that suppresses the phenotype of another interactor.'

'suppressed interactor' def: 'A interactor whose phenotype is suppressed by another interactor.'

'enhancer interactor' def: 'A interactor that enhances the phenotype of another interactor.'

'enhanced interactor' def: 'A interactor whose phenotype is enhanced by another interactor.'

'epistatic interactor' def: 'A interactor whose phenotype masks the phenotype of another interactor.'

'hypostatic interactor' def: 'A interactor whose phenotype is masked by the phenotype of another interactor.'

[pporrasebi]

Dear Chris,

I finally could put aside some time to think about these terms and I am not sure the solution we are heading towards is the correct one. My apologies for the delay, I hope this does not cause major disruption on your side.

I have realized that the terms you are requesting do not really refer to an interactor type, which in PSI-MI is defined as 'Molecular species involved in the interaction.'. In genetic interactions, the interactor type should probably always be 'gene'. The terms you are requesting rather refer to the role of these interactors in the interaction context, what we refer as 'biological role' (def: 'Physiological role of an interactor in a cell or in vivo environment, which is reproduced in the current experiment.').

I fact, I was wondering if these roles could be actually described with the current terms in the CV. The roles always refer to an effector interactor and an affected one, right? There are children of biological role 'regulator' [MI:2274] and 'regulator target' [MI:2275] that could possibly serve this role. These were created to capture causal statements and theier definitions might need some tweaking. Or new similar terms, more specific for the genetic interaction contaext, could be created.

Regarding the specific type of effect, it could be described with the rather verbose branch of the CV under 'genetic interaction' type. So this information is given at the interaction level rather than with the interactor. Have a look at the interaction type terms I propose for each of your proposed terms at the end opf the message.

So what do you think? This would hel;p keeping consistency in the way we capture interactor types and also align the representation of genetic interactions both with the original practice and with current efforts capturing causality statements.

Best regards,

Pablo.

For 'suppressor interactor' and 'suppressed interactor' [MI:0796] suppression: An effect in which two genetic perturbations, when combined, result in a phenotype that is less severe/penetrant than the most severe phenotype of the original perturbations, in effect making the organism more "wild type" in character with regards to the phenotype in question. With respect to any single quantifiable phenotype, this may be expressed as an inequality as: a < ab <= wt [E = a] OR wt <= ab < a [E = a] where 'a*' is the most severe observed phenotype value of the individual perturbations, 'ab' is the observed phenotype value of the double perturbation, 'E' is the expected phenotype value of the double perturbation, and 'wt' is the wild type phenotype value.

For 'enhancer interactor' and 'enhanced interactor' [MI:1271] enhancement: An effect in which the phenotype of one genetic perturbation is enhanced by a second perturbation to a severity/penetrance beyond (further from wild type) that expected by the superimposition or addition of effects of the individual perturbations. With respect to any single quantifiable phenotype, this may be expressed as an inequality as: ab < E < wt OR wt < E < ab where 'ab' is the observed phenotype value of the double perturbation, 'E' is the expected phenotype value of the double perturbation, and 'wt' is the wild type phenotype value.

For 'epistatic interactor' and 'hypostatic interactor' [MI:0797] epistasis: An effect in which individual perturbations of two different genes result in different mutant phenotypes, and the resulting phenotype of their combination (the double mutant) is equal to that of only one of the perturbations. With respect to any single quantifiable phenotype, this may be expressed as an inequality as: (a != wt AND b != wt AND a != b) AND (ab = a OR ab = b) where 'a' and 'b' are the observed phenotype values of the individual perturbations, 'ab' is the observed phenotype value of the double perturbation, 'wt' is the wild type phenotype value and 'a != b' indicates qualitatively or quantitatively different phenotypes.

On 30/01/2020 14:28, Chris Grove wrote:

@pporrasebi https://github.com/pporrasebi We can go ahead with the definitions proposed above:

'suppressor interactor' def: 'A interactor that suppresses the phenotype of another interactor.'

'suppressed interactor' def: 'A interactor whose phenotype is suppressed by another interactor.'

'enhancer interactor' def: 'A interactor that enhances the phenotype of another interactor.'

'enhanced interactor' def: 'A interactor whose phenotype is enhanced by another interactor.'

'epistatic interactor' def: 'A interactor whose phenotype masks the phenotype of another interactor.'

'hypostatic interactor' def: 'A interactor whose phenotype is masked by the phenotype of another interactor.'

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/HUPO-PSI/psi-mi-CV/issues/422?email_source=notifications&email_token=ADZHALY5WNE5UJW6IEFDDCLRALPZFA5CNFSM4JHLE6Y2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEKLFSQA#issuecomment-580278592, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADZHALY743AOFP7E6HDP6VDRALPZFANCNFSM4JHLE6YQ.

--

Pablo Porras Millán, Ph.D. Molecular Interactions Team Coordinator European Molecular Biology Laboratory European Bioinformatics Institute (EMBL-EBI) Wellcome Genome Campus Hinxton Cambridge CB10 1SD United Kingdom Tel: +44 1223 494482 email: pporras@ebi.ac.uk URLs: www.ebi.ac.uk/about/people/pablo-porras-millan www.ebi.ac.uk/intact www.ebi.ac.uk/complexportal

[chris-grove]

@pporrasebi Thanks.

1) The proposal was not for the proposed terms to be children of "interactor type" but rather of "experimental role". The suggested terms are not directly reflective of the biological roles of each gene or gene product, but rather an indication of the experimental role each gene or genetic perturbation plays in the experiment, similar to tagging one protein as "bait" and the other as "prey". So, it would not be appropriate to use the terms 'regulator' [MI:2274] and 'regulator target' [MI:2275]. The intention going forward is to flag the interactor type in genetic interactions with:

psi-mi:"MI:0250"(gene)

That said, with genetic interactions we are not talking about molecular species, because we are not talking about molecular interactions. We only refer to genes as interactors in genetic interactions as a sort of convenient place holder. Again, genetic interactions are not a type of molecular interaction.

2) The intention for specifying the type of genetic interaction was to use the terms in the genetic interaction branch of the CV (well, in fact, to use the new proposal for the genetic interaction branch that I'm currently working on and will submit soon as a pull request). This alone does not help us indicate which role each interactor plays. We need the ability to label each interactor with their respective role using one of the suggested terms above. Existing terms are insufficient for this purpose.

[pporrasebi]

Ah, my bad, got mixed up while following the thread. I will create the new terms then under experimental role.

One question regarding suppressor/suppressed. Does the current definition fit your use case? We might need to reconcile both to ensure backwards compatibility:

Current def on 'suppresor gene': 'Gene whose mutation suppress the phenotype associated to a suppressed mutation.' Your def: 'An interactor that suppresses the phenotype of another interactor.'

Current def on 'suppressed gene': 'Gene whose mutation phenotype is suppressed by a given suppressor mutation.' Your def: 'An interactor whose phenotype is suppressed by another interactor.'

These are not really the same. I am out of my depth when it comes to the nitty-gritty of genetic interactions, any suggestions as to how to reconcile them?

Cheers,

Pablo.

[chris-grove]

@pporrasebi OK, sounds good, thanks!

As for the 'suppressor gene' and 'suppressed gene' terms and definitions, I would prefer referencing a 'perturbation' or more specifically a 'genetic perturbation' as opposed to 'gene', but aside from that the definitions of those terms are OK except that they are quite limited by the fact that they refer to 'mutation' instead of the much broader, and more inclusive, term 'genetic perturbation'. Also, I think it is worth noting that these definitions, correctly I think, refer to the mutations as doing the suppressing or being suppressed, as opposed to the genes themselves. That notion is in-line with my earlier suggestion to make a 'genetic perturbation' interactor type as it is the genetic perturbation that does the suppressing/enhancing/masking or is being suppressed/enhanced/masked.

For backwards compatibility, we can either keep the 'suppressed gene' and 'suppressor gene' terms as they are but keep them in parallel with the new suggested terms, or we make these older terms obsolete and add a 'term-replaced-by' annotation pointing to the newer 'suppressed interactor' and 'suppressor interactor' terms.

[pporrasebi]

What about replacing the 'mutation' bit in the existing terms with 'genetic perturbation'? That would cover both use cases and should cause no backwards compatibility problems.

Tweaking a bit the definition to make it more aligned with the rest you propose, it would end up as:

'Interactor whose genetic perturbation suppress the phenotype associated to another interactor (suppressed interactor).'

[pporrasebi]

Of course, the term names would also change to the new version using 'interactor' rather than 'gene', keeping the gene ones as synonyms.

[chris-grove]

OK, after considering it more, I think I should be OK with the terms "suppressor gene" and "suppressed gene", as it will be familiar to researchers and, in some sense, isn't wrong it just uses a particular usage of the word "gene" that actually implies an allele rather than a gene in a more general sense. So, here's what I propose:

"suppressor gene" (MI:0581)

change definition from: "Gene whose mutation suppress the phenotype associated to a suppressed mutation."

to: "A gene whose genetic perturbation suppresses the phenotype resulting from a different genetic perturbation."

"suppressed gene" (MI:0582)

change definition from: "Gene whose mutation phenotype is suppressed by a given suppressor mutation."

to: "A gene whose genetic perturbation phenotype is suppressed by a different genetic perturbation."

While it is generally considered bad practice to change the definition of a term in a controlled vocabulary or ontology (as opposed to obsoleting the term and creating a new term), in these cases we are making the definitions apply more broadly, as opposed to more restrictively, so it's probably OK.

The other potential issue here, down the road, is that these definitions will restrict the use of the terms "suppressor gene" and "suppressed gene" to only those cases in which the suppressor gene suppresses a genetic perturbation (as opposed to any other perturbation, like chemical/drug/environmental perturbations) and the suppressed gene is suppressed by a genetic perturbation (as opposed to any other perturbation), respectively. I suppose we can cross that bridge when we come to it.

For the four remaining terms, then, we can propose the following:

"enhancer gene" def: "A gene whose genetic perturbation enhances the phenotype resulting from a different genetic perturbation."

"enhanced gene" def: "A gene whose genetic perturbation phenotype is enhanced by a different genetic perturbation."

"epistatic gene" def: "A gene whose genetic perturbation masks the phenotype resulting from a different genetic perturbation."

"hypostatic gene" def: "A gene whose genetic perturbation phenotype is masked by a different genetic perturbation."

[pporrasebi]

Fine with the proposal and the way forward. I would just tweak the definitions to make them a bit clearer in ths way:

"enhancer gene" def: "A gene whose genetic perturbation enhances the phenotype resulting from a genetic perturbation of a different gene."

And so on for the rest of the definitions, so it is clear that this are perturbations happening in different genes.

I will start creating/amending existing definitions and will close the issue once everything is pushed to the repo.

[chris-grove]

@pporrasebi Yes, I had initially thought of proposing the same definition (exactly, actually), but the issue is that there are instances of interactions that have been curated (at BioGRID and some at WormBase) in which the same gene is both suppressor and suppressed (intragenic suppression), for example. That's why I intentionally left off the "different gene" from the definition

[pporrasebi]

OK, will leave the proginal one you proposed then.

Pushing the commit in a few minutes, once I am done.