manulera
commented
2 years ago On negating phenotypes for rescue
Hi @ValWood and @kimrutherford. I initially thought to write this for Kim, but I think it can be a good thread to reason whether we do it / how we do it.
Example 1 - A negative annotation might be easier for community curation and the ontology
Sometimes, people will study a certain process, such as endocytosis or spindle formation, and describe a single allele phenotype that abolishes that process. For instance, cut7-24 forms monopolar spindles at high temperature, this is a major phenotype that prevents elongation of the spindle at all, and therefore chromosome segregation.
Then, they will find a double allele genotype that may prevent that phenotype from occuring. For example, in cut7-24 pkl1D the spindle is not monopolar anymore, and can elongate during mitosis and segregate the chromosomes. Often, authors will only report the rescue of bipolarity of the spindle in their paper, and nothing else, since this is a major phenotype.
In this case, I think having the NOT annotation is mostly convenient for:
- A community curator to easily annotate this genetic interaction (rescue).
- Not creating unnecesary ontology terms (see below)
The problem with this "partially normal" phenotypes is that we know that certain aspects of the normal process have been restored, but maybe not all of them, so we should probably not annotate with "normal phenotype". We could create a term that captures only a part of this normality, such as "bipolar spindle", but this cannot be a child of "normal spindle", it must be if anything a parent (since all normal spindles are bipolar, but not the other way). If we follow this, we might be creating multiple parents of "normal spindle" that are children of a very broad term (normal subcellular component). Even if we don't make them parents of the normal spindle, they would have to be children of the very generic term, despite being somewhat specific. Negating the phenotypes which are easy to place in the inheritance would make it easier.
Alternatively, I guess some special relationships may exist like part_of instead of is_a like in GO, but still one would have to replicate the full inheritance of the abnormal phenotype on the normal side instead of saying NOT.
Example 2 - A negative annotation might bring more information than conventional annotations to a double allele genotype
The authors may instead go in more detail and report phenotypes for the double mutant. In this particular example:
- The spindle that forms is shorter, so if the authors actually measured it, they could annotate the genotype cut7-24 pkl1D to "short bipolar mitotic spindle".
- The spindle may also have other defects in this double allele genotype that are contemplated in FYPO, such as "bipolar mitotic spindle with decreased, irregular thickness" (this is an invendted example).
These two phenotypes imply that the spindle is bipolar, so they could be used as a phenotype for the genetic interaction to show that bipolarity is restored.
- Single allele: cut7-24 -> monopolar spindle
- Double mutant that rescues:
- cut7-24 pkl1D -> short bipolar mitotic spindle
- cut7-24 pkl1D -> bipolar mitotic spindle with decreased, irregular thickness
If I visit the gene page of pkl1, I think the most interesting information about the double allele genotype from this study is that it rescues the monopolar spindle, so in this case I also think it would be good to have the "NOT monopolar spindle" associated to the rescue.
Example 3 - The same genotypes might be differently annotated in another study, depending on the intent or directionality of the genetic interaction
Sticking with the same example, let's imagine another study in which the authors study spindle length. They initially show that deleting pkl1 makes the spindle slightly shorter than wild-type, and that deleting cut7 on top of pkl1D makes the spindle much much shorter than wild type (this is what happens in reality). In this case, the most interesting information that comes from this study is that cut7 phenotypically enhances "short bipolar mitotic spindle", and the current associated genetic interaction dialogue would work best for this, I believe. The user could in this be case suggested to indicate penetrance / severity for both single and double genotypes to show the rescue.
Coming at this from this angle also shows that most likely most of the change in length is due to cut7D and not pkl1D, even though cut7D cannot be assayed on its own.
Finally, If cut7D pkl1D was a partial or total rescue of pkl1D (if the spindle length in cut7D pkl1D was longer than in pkl1D), I think it would make sense for the authors to indicate the partial rescue with penetrance / severity in the current dialogue. If the rescue of the length was total, maybe one could use "normal spindle length" annotation, but "NOT short spindle" is easier for the ontology probably as indicated above. I think this is the case that is less clear to me.
kimrutherford
Discussing with @manulera we would like (need) to be able to capture a genetic rescue experiment ONLY negating the phenotype of the single mutant but without necessarily specifying the phenotype of the double mutant since this does not always make sense.
In an example single mutant x has "monopolar spindle defect" allele or OverX of Y rescues "monopolar spindle defect" BUT spindle may not be "normal" We would want to do a GO type negation "monopolar spindle defect" NOT
We can discuss how and where on Tuesday.