[Biopython] Is this a valid genbank record?

[Fields, Christopher J]

On Jan 18, 2012, at 9:34 AM, Peter Cock wrote:

> On Wed, Jan 18, 2012 at 3:24 PM, Fields, Christopher J
> <cjfields at illinois.edu> wrote:
>> On Jan 18, 2012, at 8:50 AM, Peter Cock wrote:
>> 
>>> On Wed, Jan 18, 2012 at 11:11 AM, Michael Thon <mike.thon at gmail.com> wrote:
>>>> 
>>>> On Jan 18, 2012, at 12:03 PM, Peter Cock wrote:
>>>> 
>>>>> On Wed, Jan 18, 2012 at 10:14 AM, Michael Thon <mike.thon at gmail.com> wrote:
>>>>>>  They have weird feature locations:
>>>>>> 
>>>>>>     Het             join(bond(127),bond(127),bond(130),bond(130),bond(138),
>>>>>>                     bond(138),bond(139),bond(138))
>>>>>> 
>>> 
>>> Do you actually need to do anything with this feature? If not, then
>>> the pragmatic solution is we issue a warning but otherwise ignore
>>> the feature and continue parsing. I'm struggling to grok exactly
>>> what this location is trying to convey - maybe I should read the
>>> associated paper?
>> 
>> GenPept is littered with these.  With bioperl we only attempt to
>> support 'bond' types for round-tripping, but I don't recall whether
>> this has been extensively tested, though it would be easy enough
>> to add this in to see if the location factory will handle this properly
>> (both to and from a location string).
>> 
>> Do wish NCBI would document this more...
> 
> +1
>  
> 
> join(bond(127),bond(127),bond(130),bond(130),bond(138),
> bond(138),bond(139),bond(138))
> 
> rather than:
> 
> bond(127,127,130,130,138,138,139,138)
> 
> or indeed by so many of the residues are bonded more than
> once?
> 
> Peter

No, that one is particularly odd, but there isn't a reason I could see where this couldn't be supported, it's just a join of simple locations.  Seems this is something that may be auto-generated, wouldn't be surprised to see more of these.

As to whether it's a valid GenBank record, well, considering the source of the record is NCBI, I think it's safe to say it's valid.

(though again, this all comes back to how helpful it would be to have documentation re: how bond() is defined within the context of the feature table)  

chris