[go] Boundary between IMP and IGI

[Karen Christie]

Hi,

I'd like to bring the IMP IGI discussion back to the original suggestion, 
which has been lost in side issue of whether these two codes should be 
merged. I agree that they should not be merged.

In commenting on the draft of the proposed new evidence codes 
documentation, Val and Midori brought up some confusion about the use of 
IGI to annotated genes where a role is inferred based on a mutation in a 
different gene.

In a lot of ways it seems that it would be a simpler decision between IMP 
and IGI if the only question is how many genes are mutated?
 	1 ==> IMP or
 	more than 1 ==> IGI.

Right now, to make the decision, you have to ask a more complicated 
question that involves which gene is being annotated.

-Karen



On Thu, 13 Sep 2007, Michael Ashburner wrote:

> I also agree with Val, we certainly need both of these, they are very 
> different biologically.
>
> Michael
>
> On 11 Sep 2007, at 17:03, Judith Blake wrote:
>
>> I agree with Val.  With multicellular organisms, perhaps even more 
>> particularly,  it's a long trip between observable heritable phenotypes and 
>> understanding the precise genetic interactions that are producing them.
>> 
>> Judy
>> 
>> Valerie Wood wrote:
>>> Benjamin Hitz wrote:
>>> 
>>>> 
>>>> Do we really need IGI and IMP?   Is the only difference technically  that 
>>>> IGI = double (or 2+?) mutant, IPI = single mutant?
>>> 
>>> 
>>> We need.
>>> 
>>> These are *very* different types of biological data. The IMP the 
>>> annotation is derived in some way directly from the *observable 
>>> phenotype*, but with IGI it is an inference from the the actual 
>>> *interaction* (the phenotype may be suppressed or the cell may be dead).
>>> 
>>> Our current use of IGI includes things which aren't 'truly' genetic 
>>> interactions but on the whole, it is likely that most databases have 
>>> recorded non- canonical use and these can be filtered (i.e functional 
>>> complementation by a heterologous system can be filtered because the 
>>> 'with' column will contain a entry from another taxon.
>>> 
>>> Val
>>> 
>>> 
>>> 
>>> 
>>>> 
>>>> Ben
>>>> 
>>>> On Sep 10, 2007, at 4:18 PM, Karen Christie wrote:
>>>> 
>>>>> Boundary between IMP and IGI
>>>>> -------------------------------------------------------
>>>>> 
>>>>> In response to the new draft of the evidence code documentation, some
>>>>> discussion came up between Midori and Val about the usage of the IGI
>>>>> versus the IMP evidence codes. As this issue was not a specific gripe
>>>>> of anyone on the Evidence Code Committee, it was not discussed.
>>>>> 
>>>>> However, one of the goals of this revision was to have guidelines that
>>>>> make sense and I completely see the point that it doesn't really make
>>>>> sense to say that making an inference from a strain with one mutation
>>>>> is a genetic interaction, even when you are annotating a gene other
>>>>> than the one that is mutant.
>>>>> 
>>>>> We were also asked to make a decision tree/flow chart for evidence
>>>>> code decisions (I have a draft I'll send out later), and I think it
>>>>> would be a much simpler decision if there was a clear line between 1
>>>>> mutant gene and multiple mutant genes.
>>>>> 
>>>>> I think it would make a lot more sense if any annotations made on the
>>>>> basis of mutation, or comparison between alleles, of a single gene
>>>>> should use IMP.  Since we already allow use of the with field for IMP
>>>>> to record the mutant allele, it might make more sense to use IMP for
>>>>> any annotation based on a phenotype of a single gene and just record
>>>>> the mutant allele in the with field. Since not all groups track
>>>>> alleles, perhaps we should also allow with for IMP to contain the name
>>>>> of the gene without specifically designating an allele.
>>>>> 
>>>>> Below is transcript of the discussion that occurred on this issue.
>>>>> 
>>>>> -Karen
>>>>> 
>>>>> 
>>>>> **IMP:
>>>>> 
>>>>>> mutation in gene B provides information about gene A being
>>>>>> annotated. For this type of experiment, use the IGI code.  and IGI:
>>>>>> Inference about one gene drawn from the phenotype of a mutation in a
>>>>>> different gene
>>>>> 
>>>>> 
>>>>> Midori (15 Jun 2007):
>>>>>  I have always disagreed with this usage: I've argued that IMP  would be
>>>>>  more appropriate, because in the examples given, only one gene is
>>>>>  mutated, so the "combination of alterations" criterion for IGI is  not
>>>>>  met. But it's an argument that I lost years ago. Oh well.
>>>>> 
>>>>> Val (22 Jun 2007):
>>>>>  This is still a bit is unclear to me
>>>>>
>>>>>  "We also use this code for situations where a mutation in gene A
>>>>>  provides information about the function, process, or component of  gene
>>>>>  B. If a mutation in gene A causes a mislocalization of gene B,  gene A
>>>>>  is annotated to protein localization with gene B in the with/from
>>>>>  column using IGI."
>>>>>
>>>>>  In the protein localization example above a mutation in gene A is
>>>>>  providing information about gene A (protein localization) not about
>>>>>  gene B (the protein localized).
>>>>>
>>>>>  I have made a number of these type of annotations to 'protein
>>>>>  localization, (the fission yeast community are very keen on
>>>>>  localization dependency experiments for functionally connected gene
>>>>>  products). However, I thought I had used the wrong evidence code
>>>>>  (using the existing documentation) and that they should be IMP (I
>>>>>  wanted to capture the protein localized and at the time I had no  other
>>>>>  way to do it). These were on my todo list to fix.  It now seems they
>>>>>  are OK as IGI, so I just wanted to double check.........
>>>>>
>>>>>  The original documentation says:
>>>>>  # Inference about one gene drawn from the phenotype of a mutation  in a
>>>>>  different gene I don't have an example of this though. I forgot what
>>>>>  it is used for, although I used to know......
>>>>> 
>>>>> Midori (22 Jun 2007, in response to Val):
>>>>>  > I have made a number of these type of annotations to 'protein
>>>>>  > localization, (the fission yeast community are very keen on
>>>>>  > localization dependency experiments for functionally connected  gene
>>>>>  > products). However, I thought I had used the wrong evidence code
>>>>>  > (using the existing documentation) and that they should be IMP (I
>>>>>  > wanted to capture the protein localized and at the time I had no
>>>>>  > other way to do it). These were on my todo list to fix.  It now
>>>>>  > seems they are OK as IGI, so I just wanted to double  check.........
>>>>>
>>>>>  Your annotations are consistent with the existing documentation.  What
>>>>>  I'm saying is that I think the documentation should recommend IMP  for
>>>>>  these.
>>>>>
>>>>>  I think I still wouldn't put B is 'with' with IMP, because a few
>>>>>  groups would put the allele of A used in the experiment, and others
>>>>>  would leave 'with' blank.
>>>>>
>>>>>  >  The original documentation says: # Inference about one gene drawn
>>>>>  > from the phenotype of a mutation in a different gene I don't  have an
>>>>>  > example of this though. I forgot what it is used for, although I
>>>>>  > used to know......
>>>>>
>>>>>  I would also prefer to recommend IMP for these.
>>>> 
>>>> 
>>>> -- 
>>>> Ben Hitz
>>>> Senior Scientific Programmer ** Saccharomyces Genome Database ** GO 
>>>> Consortium
>>>> Stanford University ** hitz at genome.stanford.edu
>>>> 
>>>> 
>>>> 
>>>> 
>>> 
>>> 
>>> 
>> 
>