New genetic interaction type terms (BioPAX)

[noedelta]

Hi, We're starting to think about how to extend BioPAX to incorporate genetic interactions in BioPAX Level 3. We have a meeting to review this at Cold Spring Harbor June 20-21.

The proposal for BioPAX is given here: http://biopaxwiki.org/cgi-bin/moin.cgi/Genetic\_Interactions

We'd like to use the PSI-MI interaction type CV for the BioPAX 'interaction' class as much as possible. It would be nice to have some new genetic interaction terms added. A complete classification of genetic interaction types was recently published and we would like to use it to classify interactions because it formally proves that it is complete. What do you guys think about incorporating these into PSI-MI?

The following types and definitions are taken from Drees et al. Genome Biol. 2005;6(4):R38. Given phenotypes resulting from genetic perturbations: A = phenotype resulting from genetic perturbation A, and B = phenotype resulting from genetic perturbation B, AB = phenotype resulting from both genetic perturbations together, WT = wild-type phenotype, we have:

Synthetic interaction: A and B have no effect on the WT background, but the AB combination has an effect.

Asynthetic interaction: A, B, and the AB combination all have the same effect on the WT background.

Suppressive interaction: A has an effect on WT, but that effect is abolished by adding the suppressor B, which itself shows no single-mutant effect (for example, WT = B = AB < A); or, the corresponding holds under exchange of A and B.

Epistatic interaction: A and B have different effects (in terms of direction or magnitude) on the wild-type background and the double mutant has the same phenotype as either A or B (for example, A < WT < B = AB).

Conditional interaction: A has an effect only in the B background, or the B mutant has an effect only in the A background.

Additive interaction: Single-mutant effects combine to give a double-mutant effect as per WT < A= B < AB, B < WT = AB < A, WT < A < B < AB, B < WT < AB < A, and all additional inequalities obtained by interchanging A and B, or reversing the effect of both A and B.

Single-nonmonotonic interaction: B shows opposing effects in the WT and A backgrounds (for example, B > WT and AB< A); or, A shows opposing effects in the WT and B backgrounds, but not both.

Double-nonmonotonic interaction: Both A and B show opposing effects in the WT background and the background with the other mutant gene.

The following supplemental useful types and definitions are taken from the BIND database specification:

Enhancement interaction: the A perturbation enhances the phenotype of the B perturbation, or vice versa (e.g. WT = A < B < AB or WT = B < A < AB). This could be conditional or additive by the above scheme.

Intergenic non-complementation interaction: two perturbations fail to complement but act on different genes. This could indicate that both genes are physically interacting.

---

Issues: Existing PSI-MI terms that are children of genetic interaction match this classification. Their children, though, mix 2 independent concepts into the definition of genetic interaction: phenotype and mutation status. Both of these are closer to the experimental detail than to the interpretation.

1) mutation status: e.g. suppression knockout specifies that the gene is knocked out. This information is more like experimental form.

Possible types we are considering for BioPAX are: knock-out: The gene has been completely removed e.g. by genetic engineering

knock-down: The gene expression has been significantly reduced by introduction of an external substance, e.g. by RNA interference

hypomorph: The gene function has been significantly reduced by altering its sequence e.g. a temperature sensitive mutant

over-expressed: The gene expression has been significantly increased by engineering, e.g. by replacing the normal gene promoter with one that overexpresses the gene.

(and maybe chemical-target later)

2) phenotype: there are effectively infinite phenotypes, though some phenotypes are used very often to measure genetic interactions e.g. lethality. The more specific genetic interaction types, like 'synthetic-lethal', include a phenotype description (and hence experimental evidence) in the definition of the interaction type, which is not ideal because there are many phenotypes. However, this makes sense for very commonly used phenotypic experimental outputs, like lethality, because of the difficulties in formally describing phenotype. I.e. it is hard to generally define phenotype, but don't let this prevent the capture of common data types - the benefits outweigh the costs.

---

Some of the terms don't match the above classification: -conditional synthetic lethal - is this conditional or synthetic? These don't overlap by the above classification. Unless the condition term is describing another dimension of perturbation, like environmental perturbation (temperature, nutrition, like that term's children) instead of genetic perturbation.

Reported by: gbader

[noedelta]

Logged In: YES user_id=653048

Approved at the PSI meeting in Washington, terms added as children of MI:0208 'genetic interaction'

Original comment by: luisa_montecchi

[noedelta]

• status: open --> closed

Original comment by: luisa_montecchi