[BBF Standards] Representing parts

[Deepak Chandran]

Hello Raik,

As far as development of a new language, the intent was to allow people 
from different disciplines to communicate without having to cross 
hurdles. Pictures are generally the easiest way to summarize a model 
(i.e. device or systems). So, the idea was to create a format that is 
similar to what one would see in a picture and yet contain sufficient 
information. This format would be converted to other standard formats 
such as XML in the end, but it would still allow someone who is not not 
accustomed to XML to contribute. But if there are no hurdles to cross in 
the first place (which I don't know the answer to), then there is no 
need to a new language.

As far as defining interactions of parts (vs. devices), since most of 
the parts come from nature, isn't it a question of finding the 
information about these parts from literature or databases (most parts 
have probably not been "thoroughly" characterized yet, but that is an 
issue for which probably requires lots of time+money)? When dealing with 
gene-regulation, most promoters/operators can be placed in front of any 
coding region (from my knowledge), so is there a need to define 
"interactions" for these parts (other than issues such as spacing 
between promoter and RBS or codon-bias for parts from other organisms) ?

As far as representing models (devices/systems), I agree that reaction 
models are not the correct way to represent multimeric proteins (or 
other complexes). This issue comes up in gene regulation -- suppose 
there are four different transcription factors that can bind to an 
operator region, then there are 2^4 possible configurations that the 
operator region can have. For transcription, one can often assume that 
the binding/unbinding of transcription factors is much faster than 
transcription itself, which allows one to make a single rate expression 
for transcription based on the concentrations of the four transcription 
factors. That is possible only because transcription processes are 
generally much slower than protein-DNA interactions, which does not 
apply to multimers. However, even with a system of rules, isn't the 
final model still in terms of reactions, although the model itself would 
not contain all the combinations? 

--Deepak

Raik Gruenberg wrote:
> Hi Deepak,
>
> it looks like we should have a dedicated section about the simulation 
> of biobrick networks. There is obviously a large interest here. IMO 
> your format is perhaps more suited for the exchange of device models 
> rather than for the nuts and bolts specifications of a biobrick. 
> Models *are* an interesting problem, although right now, I am a bit 
> more concerned about getting the basics sorted out A.S.A.P because we 
> otherwise may end up not having any real biobricks to model :-/. IMO, 
> the model exchange needs more research and practice before we can nail 
> down standards -- this is not really my area but here are some hints 
> in that direction:
>
> Reaction models (also SBML) have a fundamental flaw -- they cannot 
> cope with the combinatorial complexity that is common to signaling. 
> For example, picture a hexameric protein receptor built from 3 
> different monomers a, b and c with shifting abundance and different 
> affinities for each other. This could yield 729 possible 
> receptor-subtypes (aabbcc, aaabbb, ababac, ...) each with different 
> specificities or affinities to it's ligand(s). Now try to write down 
> the reaction model of that system! And this is only the beginning of a 
> perfectly normal signaling pathway... This may be a problem you don't 
> have with true PoPs devices but it certainly comes up for protein 
> devices.
>
> An alternative to reaction models are rule-based models. Please have a 
> look at this paper:
> Hlavacek WS et al. (2006) Rules for modeling signal-transduction systems.
> PMID: 16849649
> I think we should rather exchange rules than complete models. Biobrick 
> 'a' would only need an 'interaction+affinity' link to biobrick 'b' and 
> 'c' and the hexameric receptor could be described as device assembled 
> from 6 times a, b or c. The large reaction model can then be built 
> fully automatically and additional information can be incorporated as 
> it arrives.
>
> Another advantage, those kind of rules can be expressed by simple 
> relations between biobricks (and devices). There would be no need for 
> a new language and we can start collecting these basic rules 
> (interaction-, activity data) already now.
>
> Also I have some reservations about creating a completely new format / 
> language because it tends to seal off a community rather than opening 
> it up. But there is also the opposite school of thought and new 
> problems are perhaps sometimes indeed best tackled with new languages...
>
> Greetings,
> Raik
>
> Deepak Chandran wrote:
>> I had written some time back about implementing a simple file format 
>> to represent biological parts. I apologize for taking so long to get 
>> back to that. Anyway, the following pdf has a description of what I 
>> (we) am proposing:
>>
>> http://www.washington.edu/staff/deepakc/PartSyntax.pdf
>>
>> This is just a working idea. Hence, I welcome any criticisms, 
>> complements, complaints, etc.etc. In order to test the language, I 
>> wrote a small scripting program that can "connect" models together 
>> and simulate them. The link in given at the end of the pdf.
>>
>> _______________________________________________
>> Standards mailing list
>> Standards at biobricks.org
>> http://biobricks.org/mailman/listinfo/standards_biobricks.org
>>
>>
>