Talk:Synthetic Biology:BioBricks/Standardization

JM, 8/20/06

This came up earlier when we were starting out with the screening plasmid (which unfortunately doesn't work all that well for characterizing parts). If we trust our measurements (we think that they accurately reflect the system behavior) then I definitely think that it's better to start characterizing parts now. As mentioned below, all our measurements right now are going to be relative - let's pick the best standard we can think of now and, if things change in the future then we just recalibrate (c.f. changing atm->bar).

The important question, for me, is whether our measurements are valid, and here I think the data is inconclusive. Someone (and I'm not volunteering) would need to show that the measurements are both internally consistent (if A is twice as strong as B, and C is twice as strong as A, then C should be four times as strong as B) and externally consistent (if A is twice as strong as B when GFP is used as the reporter, it should also be twice as strong when RFP is used as a reporter). As long as someone's willing to do this for the 'standard composite' then I think it would be really valuable.

I think this is really the first issue that needs to be settled. Once you have consistent measurements in a completely standardized situation, then we can start figuring how the measurements map onto a new set of conditions. If you don't trust the measurements in the first place, though, trying to compare measurements in different conditions is going to be a mess. -JM

---

--JCAnderson 16:48, 20 August 2006 (EDT)

It doesn't sound like we are going to fully agree on a specific standard. It is my belief that the "standard composite" is an entirely arbitrary reference point as long as the choice doesn't in hindsight turn out to be an oddball.

For now, it seems like we do agree on several points:

• • Eventually one "standard composite" should be defined, and all activities can be a reference to that
  • The Neidhart media is preferrable to LB to minimize media-based variation
  • A single-copy standard would be ideal, but we do not currently have an appropriate FP and single-copy plasmid to do it

I still believe we should pick a temporary standard to get the ball rolling, it should reflect the content of the current registry rather than trying to anticipate what people will use in the future, and it should be as easy to use as possible. It very well may be true that a high-copy standard poorly maps to single copy experiments. We don't yet know that's true.

The real heart of the problem is not what the standard is but rather what are the mapping functions that allow us to predict the behavior of a part in one context versus another. How many of these things (changing to a slower growing strain, changing the copy number of a plasmid, changing the media, changing the temperature) result in simple scalar multiples of the activity of the part measured in a different context. In essence, the initial question that needs to be addressed is what are the sources of variation beyond the ones we already know about. How common and significant are various types of context effects, what are these various effects, and how can we eliminate them.

the Modularity problem with promoters

Drew's example of the promoter and ribosome binding site that produce a hybrid 5'UTR hairpin that disrupts translation is a separate standardization issue. I think it is more pressing, though.

The site: http://parts.mit.edu/registry/index.php/Help:BioBrick_Prefix_and_Suffix under the BioBrick Prefix section has a really critical piece of information on how to design biobrick basic parts, and I think we should add to that a preferred way of biobricking the promoter initiation site relative to the polylinker to avoid heterogeneity 5' to the biobrick junction. Again, it is an arbitrary standard, and the options are (with the transcription start in bold): Define it like r0040(and what iGEM2006 did for the family of constitutive promoters):

...ctACTAGT

Or have it in the biobrick site explicitly, something like:

...ACTAGT

So that nothing has to be re-made, and so that more native promoter sequence can be present in the part I lean towards defining the standard as the r0040-compatible version.

Clearly not all promoters are going to be compatible with this standard. Some promoters have operators that overlap or extend beyond the transcriptional start. When making basic promoter parts, one has to currently make an arbitrary decision as to where to put the 3' end of the promoter. It would be preferrable to have a standard.

On a second point, this information is really buried on the registry, but it is really important. I only became aware of the different standard for ORF parts when I was examining sequence data and noticed that they needed the start codon to overlap the XbaI site. Until that, I thought all parts were supposed to be the same way. The berkeley iGEM students last year biobricked their open reading frame and lock(rbs) parts incorrectly according to this standard, so I'm not the only one who missed it. This information needs to be more clearly declared on the registry "add a basic part" page.

On Aug 17, 2006, at 11:58 AM, John Anderson wrote:

Thanks for your comments, I'm glad we're having this discussion.

On 8/17/06, Barry Canton wrote:

I agree with Drew and Reshma's comment. Three additional comments -

1.) We might want to specify a couple of reference constructs so that people can confirm the linearity of whatever measurement system they use. One point on the curve is very useful, 2-3 might be better again.

• Chris: I agree. At the heart of it I think one construct needs to be the "absolute standard composite" and has all the parts which become defined as activity=1. Everything, regardless of the method of measurement, can be described relative to that standard composite. So, to standardize promoters, you would hold rbs, FP, term the same and vary promoter. To standardize rbs, you'd hold promoter, FP, term the same and vary just the rbs. Then, of course, you'd want to build double perturbation variants and show that their function is predicted from the single perturbation standards. At the end of the standardization, I agree you'd want a panel of variants for each style part so that the measurement for the new part can be done in parallel with the standard panel.
• Reshma 17:57, 17 August 2006 (EDT): Agreed.

2.) While I'm in favor of just picking something and then sticking with it, we might want to wait until we get one of the new FPs (emerald?) biobricked since it should be much brighter than our current standard, E0040 and would allow us to work at much lower copy. Jason has been working on this FP and might have more to add.

• Chris: If the delay for having a functioning better variant of the FP is fairly short, I'm all for it. It probably would be good to include the LVA/LAA tag on it in the standard.
• Reshma 17:57, 17 August 2006 (EDT): I am not sure about including an LVA/LAA tag. In my hands it often reduces fluorescence levels sufficiently that it can be hard to measure at low-medium copy number or for reporters other than GFP. I prefer to omit the tag.

3.) I'd like to pick a reference construct right now but the real benefit of such a construct would probably be felt by next year's iGEM teams, its probably too late for most of this year's teams to standardize. So it seems like we have the time to make a few improvements to the standard (different promoter as John suggests, maybe new FP etc.). We can gain experience with it ourselves over the winter so that it can be distributed as a well tested reference construct to iGEM next year.

• Chris: It's definitely too late for this year's teams to standardize things. I'm not sure that the teams are the ones that will be using this at first anyway. The standardization needs to begin with very simple parts--rbs, FPs, terminators, and simple promoters (constitutive promoters and operons with all trans-acting elements encoded on the genome). I doubt any iGEM teams (other than us) will have focused on those types of parts.
  There isn't a rush to standardize, but I think it is important to have a standard defined if people want to standardize, and also valuable to make an attempt to standardize the more popular parts, or introduce a set of new standardized parts.

Excited by the prospect of coast-coast standardization.

Barry

On 8/17/06, Reshma P. Shetty wrote:

Hey Chris and Drew,

Drew forwarded on your email regarding standard characterization conditions. We've definitely thought about this some. Some quick thoughts below.

See

http://openwetware.org/wiki/Synthetic_Biology:BioBricks/Standardization

and associated pages linked from there.

In particular our discussions on standard operating conditions have differed from your choices on the following counts ...

1) media -> defined rich media like Neidhardt EZ Rich media (LB tends to vary a lot batch to batch) http://openwetware.org/wiki/Synthetic_Biology:Media

• Chris: Neidhard it is.

2) strain -> MG1655 or derivative thereof (or really Blattner's strain but since no one can get ahold of it and it is not redistributable ... not very practical). I don't think DH10B is sequenced. http://openwetware.org/wiki/Standard_E._coli_Strain_for_BioBricks

• Chris: Not sure I agree with that one. On the one hand a sequenced genome is a big advantage since you know exactly what's in it. On the other hand, MG1655/W3110 is a pain to work with--it doesn't transform very well, it's recA and endA positive, etc. I don't think my iGEM team could handle MG1655. I often have difficulties working with it myself. DH10B is partially sequenced; contigs of most of it are on the web, but it isn't completed. DH10B and it's homologs and derivatives (GeneHogs, invalpha, etc.) are the most popular strain in use today and is very user friendly. Also, I think "having been sequenced" is a red herring. In probably a year or two DH10B will be finished, as will sequences of most other strains.
  In the end, I think it is going to be arbitrary what strain you pick as long as the fundamental standard is relational rather than absolute. Here, the activity of all parts is defined as its activity as a substitution in the absolute standard composite relative to the absolute standard composite's activity--independent of measurment strategy. So, that could be western-based PoPS and RIPS values, QPCR, Tecan measurments, FACS, or whatever else. In practice, we need to do all these measurements and establish 1) what are the types and sources of variation, and 2) what methods of measurment produce similar standard values. I think what you will find is that a q. Western and a FACS experiment give the exact same standards, and they will correlate well with QPCR and Tecan data when just accounting for nonlinearity.
  Once you have a set of standards defined in an arbitrary reference strain, you can take the panel of variants and probe them in a MG1655, the new Blattner strain, or whatever other strain and ask whether the relational standards hold up.
  The bottom line though is that most people are going to DO their experiment in DH10B and are therefore best served by a set of standards that will apply to their experiment.
• Reshma 17:57, 17 August 2006 (EDT): The main thing that MG1655 has going for it is that it was in some sense chosen by the E. coli as the "standard" K12 strain. That's why it was the one sequenced. Hence, it is as close to standard E. coli as there is. In my mind, choosing any other strain is pretty arbitrary. (Granted that often standards are picked somewhat arbitrarily).
  Also, keep in mind that we are just talking about a strain used for characterization purposes. Systems can be assembled in other, friendlier strains and then moved to MG1655 in the final step for characterization. Therefore, it shouldn't be so much of a burden to transform intact plasmid into it once for characteriz