Endy:Measkit PLO

From OpenWetWare
Jump to navigationJump to search

Endy:Measkit draft


  1. Engineering many-component systems is made easier by developing collections of standard parts that can be reused.
  2. It is easier still to predict the behavior of engineered biological systems assembled from standard parts if the component parts themselves were well characterized.
    • be scholarly here, making reference to current parts collections, and lack of characterization thereof.
      • For example the registry of biological parts contains X promoters and Y RBSs Z of which have been measured.
  3. Measurement of physical objects is well understood and has been successfully developed and applied in other domains (e.g., principle of correlation, et cetera). At least 3 lessons:
    1. In order to make a measurement we need (1) a method of measurement and (2) a principle of correlation. For example... in measuring length...
    2. To make reproducible measurements we should calibrate instruments and control for conditions (or used fixed conditions). Example...
    3. When you can't keep conditions constant you can determine a model for the effect of changes in conditions on the quantity you are measuring. (e.g. coefficient of thermal expansion) Example...
  4. Measuring biological parts consistently has proven challenging, and may be unlike past experiences (due to...)
    • Due to the sensitivity and unpredictability of part performance to measurement conditions (e.g. cell strain)
    • "There is no such thing as a standard component, because even a standard component works differently depending on the environment." -- New York Times, Tuesday, Jan 17, 2006, Custom-Made Microbes, at Your Service
    • Pathology / Radiology - need for experts in these disciplines, exact fit?
  5. Nevertheless, could we and would it be useful to frame measurement problem in context of past lessons, and could these be reused or adapted to impact variation in instruments and conditions, et cetera.
    1. For example, if it were the case that we hoped to actually characterize activity by the number of polymerases by using the GFP synthesis rate, would it be possible to make a model that relates GFP synthesis rates to PoPS, providing a principle of correlation for our measurement approach?
    2. Can we calibrate instruments to allow reliable reporting in PoPS? (e.g. multi-instrument calibration) (We did not test this experimentally.) Should we expect that researchers can utilize consistent operating conditions across many experiments?
    3. Would it be possible to develop a model, at first empirical, that enables prediction of promoter activity in response to varying physical conditions? For example, coefficients of (thermal) expansion...
  6. Until we address these questions in a rigorous way, we won't know if the situation is as hopeless as one might imagine based on paragraph 4 (e.g. biology can't be standardized).
    • Until we do X we cant figure out Y
    • Until we distribute and make use of calibration standards we can't report promoter activities in equivalent units.
    • Until promoters are consistently related to a reference standard we can't determine if we can build a predictive model of promoter activity in different conditions.
    • Until we define the relative strength of many promoters across many conditions we won't be able to know if predictable re-usability of components is possible in biological engineering.
    • Important Link
  7. Thus, to try these ideas out / begin to make progress, we designed reference standards for promoters and RBSs, and developed models, that taken together allow for (accounting of some sorts of variation).
    • We did the following (i) used a reference standard to evaluate variation in conditions and instruments ourselves, and (ii) distributed a reference kit to validate the approach across multiple labs. Taken together we demonstrated the utility of the reference standards (prefatory summary here).

Results (1,2, and 3 roughly correspond to 3.1-3 and 5.1-3 above)

  1. To enable measurement of promoter activity in PoPS via detection of GFP fluorescence we established a model relating GFP synthesis rate to PoPS.
    • We are unable to measure PoPS directly. Thus we must be able to relate a measurable quantity (such as changes in GFP fluorescence) to the quantity we want to measure (PoPS).
    • The relationship between GFP synthesis and PoPS is too complicated to simply intuit, so we built a quantitative model relating promoter (PoPS) and RBS (RIPS) activity in terms of GFP synthesis rate. E.g. principle of correlation for our GFP-expression based measurement approach.
  2. With our principle of correlation and measurement approach in hand, we are able to measure the same promoter across many conditions to evaluate how serious a problem we might have with part activity changing due to conditions.
    • We measured two promoters across 9 conditions and reported the results in PoPS.
      • Convert ‘GFP synthesis’ levels to PoPS via our models across different conditions to provide a picture of the range of variation in PoPS (need to do this will require some data analysis).
    • Varying conditions can result in significant changes in the activity of promoters (PoPS).
  3. Given that there is a lot of variation (between fairly similar conditions!), you might imagine 2 approaches for enabling reliable reuse of components : (1) use parts in identical conditions to those in which they were measured or (2) building models to predict changes in activity across varying conditions.
    • Both approaches are challenging. It is unrealistic to expect that researchers will be able to use standard conditions in the short term. This is both due to practical constraints of replicating the exact media, cells, growth conditions, etc, in independent laboratories as well as due to engineering constrains where different applications simply require different operating conditions (much as cars must drive in different temperature and weather). Thus approach (1) seems infeasible at this time.
    • Building theoretical models to predict changes in activity in response to conditions is complicated by the complexity of the 'biological conditions' in which parts operate. Furthermore, we lack the extensive quantitative data across many conditions that would be needed to build such models. However, in the short term we may be able to create empirical models for converting between different conditions by measuring the same promoter across multiple conditions. It is an open question how good these models will be? Will they hold across many different promoters? (e.g. if one promoter has 2X expression under condition Y does another?)
    • Thus we did the following. Show data converting measurement of promoter in SRUs across 9 conditions.
    • These results suggest that using a reference promoter may enable us to control for variation across some set of measurement conditions (e.g. varying strain, plasmid, temperature, media). It seems that across some set of standard? should there be a name for these? conditions the relative activity of at least two promoters may remain fairly constant.
  4. We may also adopt a strategy where in some cases we can take a ‘shortcut?’ and only measure a reference promoter without calibrating the instrument to allow for measurement in PoPS. Some instruments may be hard to calibrate to real GFP in first place (e.g. FACS machines) others are easier (microscope, plate reader).
    • In this case we still control for conditions and since we are taking a relative measurement calibrating the machine is unnecessary as long as both the test promoter and the reference promoter are measured by the same machine under the same conditions.
    • Show figure of lab-lab variation across 6 schools with a set of 4 promoters. Report the coefficient of variation in the mean relative promoter activity.
  5. Reporting measurements in relative promoter units (RPUs) will allow for characterization of activity that is comparable across some range of conditions and thus is a valuable characteristic to report when describing a part in such a way that it can be re-used.
    • To support part re-use we measured a total of 15 promoters and RBSs to bootstrap a collection of components characterized in this way.
    • We also listed a set of (again, need a name?) conditions across which a test promoter set maintained fixed RPU measurements within some range of error. (Obvious opportunity for future work here to (1) define a test set of promoters and (2) specify allowable error for a new condition to be added to the OK list).


  1. Reference standards are good because allows for a consistent framework in which to collect data that will support (first) empirical models regarding how activity of promoters and RBSs varies across conditions.
  2. Reference standards are also good because theory of biological engineering not yet developed, and an auto-scaling / relative measurement may be more useful for many applications.
    • make use of counter example of importance of OFF being OFF, for some applications
  3. Future Work
    • PoPS
      • Specify the bounds on the unit scale based on different measurement approaches (e.g. what is the lowest PoPS that can reliably be detected? what is the highest?)
      • What is the intrinsic error for different measurement approaches? (e.g. if you measure the same promoter many times how much does the measurement vary?)
    • SRU
      • Determine the set of experimental conditions in which the SRU holds.
      • Specify a set of test promoters for evaluating new conditions (should probably be more than 2 and should span a wide range of SRUs)
      • Make a recommendation for the error that should be tolerated for a condition to be included as "SRU-approved"
    • Improved measurement approaches
      • Measurement approaches that get closer to measuring PoPS directly.
        • Note that GFP synthesis is several steps (transcription, mRNA degradation, translation) downstream from promoter activity (PoPS), so it is necessary to use a fairly complicated model with several experimentally determined parameters. As a result, it is a non-trivial burden on researchers to use this measurement approach.
    • Validation of the GFP->PoPS model