Search Mailing List Archives

Limit search to: Subject & Body Subject Author
Sort by: Reverse Sort
Limit to: All This Week Last Week This Month Last Month
Select Date Range     through    

'with' field contents for IPI annotations

Evelyn Camon camon at
Wed Mar 1 03:19:05 PST 2006


I had a discussion last August with GO tops
about the identifiers in with 'with' field for IPI and IGI codes.

For IGI the decision was that when GOA integrates IGI data we should NOT 
convert the identifiers to protein the inference is from 
genetic data and it would be an inference too far to say that this 
interaction is at the level of the interacting gene's protein (it could 
well be at the RNA level). We integrate these as they come..but we do 
have the pipe issue still to resolve internally.

For IPI the decision at the time was that it was ok to convert 
identifiers provided to UniProt protein ids..but now the suggestion 
appears to be that we should not convert identifiers here too....????

At the moment all IPI data generated directly by UniProt/GOA staff will 
have a 'with' column filled in. Talking with GOA staff just now we quite 
liked that restriction. If the exact protein is unknown a general term 
is requested or used such as 'actin binding' or
the annotation added as IDA (no with) instead of IPI...which admittedly 
was a bit of a fudge or the anotation was left out and another paper 
with better evidence sought.

At the moment I believe we are only integrating IPI data which has 
UniProt Ids e.g MGI. We know we need to allow other identifiers to 
appear but if a protein id can be put in the 'with' for IPI data then 
please please provide it..otherwise we have on our hands an identifier 
nightmare at this end having to integrate from so many different sources.


Suzanna Lewis wrote:
> I'm with you on this Karen, but apparently it made it into the 
> documentation somewhere.
> On Feb 28, 2006, at 5:01 PM, Karen Christie wrote:
>> I don't believe we've ever implemented a restriction that the ID in the
>> with column had to be a protein ID for IPI with, because I would have
>> objected strenously to the suggestion, especially early on when some SGD
>> genes didn't even have corresponding protein IDs.
>> -Karen
>> On Tue, 28 Feb 2006, Suzanna Lewis wrote:
>>> BTW, I'm not sure when and where that restriction on the with column
>>> arose, it seems quite unnecessary.
>>> -S
>>> On Feb 28, 2006, at 3:43 PM, Doug howe wrote:
>>>> Perhaps this also relates to the relationship between the evidence
>>>> code and the allowed type of object in the 'with' column.  Does it
>>>> create a computational problem if IPI annotations have both protein
>>>> and gene IDs in the 'with' column since one is sequence and one is DB
>>>> identifier?  I suppose it isn't all that bad since no matter what you
>>>> put there, it will have a dbxref associated with it...
>>>> -Doug
>>>> Kimberly Van Auken wrote:
>>>>> Hi Doug--
>>>>> We have come across this situation here at WormBase and have found
>>>>> that even though we have
>>>>> protein IDs corresponding to all confirmed (or even partially
>>>>> confirmed) transcripts, it is often
>>>>> difficult, if not impossible, to tell exactly which proteins authors
>>>>> studied in interaction experiments,
>>>>> be it two-hybrid assays, co-IPs, etc.
>>>>> In these cases, we feel that putting the gene ID in the WITH column
>>>>> is the best compromise.
>>>>> Doing so still allows users to identify the interacting entities, but
>>>>> does not make any potentially
>>>>> misleading statements about exactly which proteins were involved.
>>>>> I am also curious to know how other groups handle this. Best,
>>>>> Kimberly
>>>>> Doug howe wrote:
>>>>>> We (curators working from primary literature) often find
>>>>>> circumstances for IPI annotation where the authors do not provide
>>>>>> specific information about exactly which protein they are working
>>>>>> with...though we usually know which gene it came from.  So when
>>>>>> making IPI annotations, where the exact sequence of the protein
>>>>>> involved is questionable, is it better to put a representative
>>>>>> protein ID in the "with" field, put the ID of the gene itself in the
>>>>>> with field (is this even valid?), or leave the "with" field empty?
>>>>>> How often do other groups actually find that they know the exact
>>>>>> protein sequence ID of the interacting proteins?
>>>>>> -Doug

Evelyn Camon
GOA Coordinator
Senior Scientific Curator
European Bioinformatics Institute
E-mail: camon at

More information about the go-discuss mailing list