Advertisement
Research Article

Estimation and Discrimination of Stochastic Biochemical Circuits from Time-Lapse Microscopy Data

  • David Thorsley mail,

    thorsley@u.washington.edu

    Affiliation: Department of Defense Biotechnology High Performance Computing Software Applications Institute, Telemedicine and Advanced Technology Research Center, U.S. Army Medical Research and Materiel Command, Fort Detrick, Maryland, United States of America

    X
  • Eric Klavins

    Affiliation: Department of Electrical Engineering, University of Washington, Seattle, Washington, United States of America

    X
  • Published: November 06, 2012
  • DOI: 10.1371/journal.pone.0047151
  • Published in PLOS ONE

Reader Comments (2)

Post a new comment on this article

From the peer-review process

Posted by jpeccoud on 12 Nov 2012 at 13:43 GMT

Below is the author's response to the reviews of the original submission. I hope it can help place the paper in perspective.
-------------------------------------------------------

"1- Title and headings: I agree with Reviewer 2 that the choice of your heading is
somewhat misleading. Your algorithm seems to be capable of discriminating
between two models but it would not be practical to estimate unknown parameter
values. Similarly the title of the paper should be changed to better reflect the
results of your work."

We have changed the title to “Estimation and Discrimination of Stochastic Biochemical Circuits
From Time-Lapse Microscopy Data.” Our algorithm allows for a coarse-grained parameter
estimation by allowing the models to differ not in structure, but only in parameter values, but this
situation is, as the Reviewer and Editor argue, more accurately described as model
discrimination. Ideally, the short title for the paper would demonstrate that our approach can do
both state estimation and model discrimination, but given the choice of one of the other, we will
follow your advice and use the heading “Discrimination of Stochastic Biochemical Circuits.”


"2-Synthetic Biology: I agree that your work could be potentially useful in
synthetic biology but you seem to take for granted that synthetic biologist already
work with well parameterized models. I suggest that you develop this point a little
further by citing synthetic biology references that include well-parameterized
models (there are not so many of them) and/or discuss if synthetic biology will
actually be able to develop such models. I recently published a review in Trends
in Biotechnology about Genetic Design Automation where we discuss this
problem. Feel free to cite it if you think this is useful."

We have toned down our statement on the current ability to parameterize synthetic biology
models. We feel it is not necessary to say anything more than that synthetic biology models tend
to be smaller and thus have fewer parameters to estimate.

"3- As pointed out by Reviewer 2, fluorescent proteins themselves have a number
of limitations that make it problematic to reconcile time series of fluorescence
data with the dynamics of the underlying model. See for instance the paper we
published in PLOS ONE last year: Oscillatory Dynamics of Cell Cycle Proteins in
Single Yeast Cells Analyzed by Imaging Cytometry. You may want to discuss it."

We revised the discussion to make it clear that the three limitations we discuss are the not the
only issues with fluorescent proteins and cited the recommended paper so that the interested
reader can learn more about this issue. We also state clearly that the need to parameterize the
sensory apparatus is a drawback to our approach.

"The algorithm described in your manuscript is timely and interesting but the
overly optimistic tone of your paper is likely to give readers that it does not
deliver on its promise. If you tone it down, the impact will be increased."

We are excited about this paper because it presents a general theoretical solution to the problems
of state estimation and model discrimination for stochastic chemical systems. The trade-off to
this generality is that computation the solution to the problem using the method described is
typically not tractable. Our hope is that the mathematically-interested biologist, or the
biologically-interested mathematician, will be able to use this general solution as a starting point
to derive more tractable solutions for specific classes of models. We hope that the revision has
made it clear that we do not expect immediate implementation of our method to solve biological
problems, but that these problems can be solved with appropriate domain-specific extensions of
this work.

No competing interests declared.