Biology Reference
In-Depth Information
initial set of only 20 genetic variants out of 125 (5
3
) combinations. The central composite
model is particularly useful for response surface methods,
56
and essentially fits a second
order polynomial model to the observed productivity as a function of variables similar to
the one below:
X
n
X
n
X
n
1
i
X
i
1
Y
1
β
i
X
i
1
j
β
ij
X
i
X
j
5β
0
1
1
β
i
i
1
i
5
5
#
,
where
Y
is the response,
X
1
...
n
the factors,
β
0
a constant,
β
1
...
n
and
β'
1
...
n
model coefficients
for the first- and second-order factors and
β
12
...
(
n
2
1)
n
the parameter for the interaction
between variables. This type of design is particularly important when one wants to evaluate
the possibility of nonlinear responses and interactions between variables, which is often the
case in genetic circuits.
DoE effectively reduces the overall number of experiments without compromising the
quality of the data. In addition, it enables a close integration between the design and analysis
phases of an experiment such that models derived in the analysis phase can feed back into
the next round of experiments to be designed (
Fig. 4.4
). We believe this strategy will take a
central role as our ability to model circuits accurately and the number of parts necessary for
perturbing them grows.
For both directed evolution and DoE strategies, the ability to vary circuit parameters in
efficient and predictable ways, thus creating rationally targeted and distributed variation, is
crucial. Therefore, these methods rely fundamentally on the availability of well-characterized
parts. In addition, they become only more powerful as part assembly becomes more
predictable, since the more reliably one can target relevant regions of functional space,
the faster directed evolution and DoE strategies can hone in on exact behavior.
CONCLUSION
Synthetic biology aims first to develop modular biological parts and then to assemble
them into composite function.
5
There is an increasing number of parts that can now be
easily designed and tuned to achieve desired functionality using models that describe their
behavior. Additionally, there are emerging paradigms that ensure that parts can be
composed reliably to yield predictable behavior. Together, these elements provide a more
comprehensive representation of a part datasheet,
57,58
and are starting to move synthetic
circuit design towards a more computer-aided design (CAD) framework.
59
However, for
the foreseeable future there will always be significant uncertainty in part function and
interaction with the host and other components. These issues of parasitic interaction and
load, including competition for molecular resources and unpredictable chemical toxicity,
currently limit an entirely rational approach to circuit design.
76
Instead, we have argued in favor of an integrated approach for circuit design consisting
of parts design, assembly and modeling, and the use of well-characterized parts as sources
for systematic genetic variation for two distinct search methodologies: directed evolution
and DoE (
Fig. 4.2
). The former has the exceptional ability to quickly generate millions of
variants, but strongly depends on availability of high-throughput screening methods. On the
other hand, DoE offers a very flexible strategy to minimize experimental effort and provide
learning about system dynamics. This integrated approach should result in efficient design
cycles for the creation of custom biological function and a rich knowledge base for
interaction and load that can ultimately inform the design of future applications.
Acknowledgments
JCG acknowledges financial support by the Portuguese Fundaçãopara a Ciência e a Tecnologia (SFRH/BD/47819/
2008). This work was supported by the Synthetic Biology Engineering Research Center under NSF grant number
0540879 and by the BIOFAB under NSF grant number 0946510.
