CAGEN: Robust Gene Response Challenge

2012 CAGEN Challenge Problem

This challenge problem has been chosen as the 2012 CAGEN Challenge Problem. To participate in the 2012 CAGEN Challenge:

1. (optional) Send e-mail to the chair of the CAGEN steering committee (murray-at-cds-dot-caltech-dot-edu), indicating that you plan on participating in the challenge problem
2. Subscribe to the cagen-announce mailing list to receive information about the CAGEN challenge
3. Submit a technical paper describing the results of your design by the deadline for the competition (15 June 2012)

More information on the CAGEN challenge motivation and principles is available on the CAGEN home page.

WARNING: THIS CHALLENGE PROBLEM IS STILL IN DRAFT FORM This CAGEN challenge problem is still in draft form and may be revised. For more information, see the CAGEN home page.

Challenge Problem Description

Sample step response, taken from http://partsregistry.org/Part:BBa_F2620:Response_time.

Synopsis: The goal of this challenge is to design a circuit that can express a fluorescent protein at a controlled level upon the introduction of a chemical inducer, with minimal variation in expression between cells and in multiple contexts. At conditions yielding maximum expression, the circuit should quickly bring the volume-normalized fluorescence from 1X to 10X in response to the addition of an inducer of the designer's choice. The circuit must work at multiple temperatures, with minimal variation in the fluorescence over time, operating temperature and cell choice.

Motivation: Current synthetic circuits demonstrate large variability in expression level when operating different contexts and this limits the ability of synthetic biologists to build on designs performed by other groups. By designing circuits that demonstrate highly repeatable performance over a range of operating conditions, it will be possible to make better use of designs in a modular fashion.

Impact: Improved understanding engineering processes for synthetic biologists will enable more rapid and pervasive development of synthetic circuits, with applications in materials processing, environmental science, agriculture and and medicine.

Metric(s): The winner of this challenge will be determined based on the worst case, mean square error between the ideal step response and the experimental results, with evaluation over multiple temperatures. To be considered, data for the circuit must be submitted for steady state operating temperatures at a nominal value (chosen by the contestant), nominal + 5% and nominal - 5%, with measurements taken in at least 5 individual cells chosen from separate colonies. This represents a set of 15 total time traces of data. At least one of these responses must demonstrate a step response that goes from 1X to 10X expression level in response to the addition of the inducer.

The following method will be used to determine the numerical score for each time response: let [math]\displaystyle{ r_i(t) }[/math] represents the (single) ideal step response at a given induction level i (with minimum and maximum values chosen by the participant), [math]\displaystyle{ y_i(t) }[/math] represents the measured fluorescence of a given cell, T1 represents the time at y(t) reaches 5% of its maximum value and T2 represents that point at which it reaches 95%. Each run will be scored according to the formula:

Score[run] = [math]\displaystyle{ \int_{T1}^{T2} |y_i(t) - r_i(t)|^2 dt }[/math]

The score for the submitted design will be the worst (highest) value of the score across all runs (15 total). Note that [math]\displaystyle{ r_i(t) }[/math] is fixed based on induction level, while [math]\displaystyle{ y_i(t) }[/math] depends on the specific run. The participant can specify a single minimum value for [math]\displaystyle{ r_i(t) }[/math] and a table of maximum values (one for each induction level).

Contact: To provide feedback on this challenge, send e-mail to Richard Murray (murray-at-caltech-dot-edu), representing the steering committee.

Benchmark Data

Test implementations of baseline circuits in E. coli and S. Cerivisiea" are being built and characterized. Benchmark data and scripts for computing the score will be posted here over the summer of 2011.

Alternative metric to calculate the performance score

An alternative to the proposed metric is to calculate deviation of trajectories from an ideal step response. Red trace denotes a trace chosen as reference. Shaded grey regions denote areas penalized by the two metrics. For the proposed metric figure, this is the region where trajectories lie. For the alternate metric figure, this is the region separating trajectories from an ideal step response.

The current metric aims to capture the performance specifications by calculating the worst-case error among traces. In principle, there is another way to calculate the performance score. This is to estimate the deviation of measured traces and an ideal step response. The difference between these two metrics is illustrated in the adjoining figure.

Tools for the Robust Gene Response Challenge

The specifications of the Robust Gene Response Challenge are to design a genetic circuit that ensures fast, robust expression of a fluorescent protein upon induction. For both metrics, we have worked through the specifications for a reference design to establish a baseline performance score, using both computational and experimental methods. This analysis has helped refine the specifications and will provide tools and protocols to help participating teams in their designs.

For the original metric, this analysis is presented in the following report (pdf), and also in the following presentation charts (ppt, pdf, movie1, movie2). The associated latex files, scripts, and data can be accessed here. Similar data has also been acquired using a microfluidic setup (CellASIC ONIX), which offers better control over the induction time and the possibility of pulsed induction (Movie, Preliminary Data).

For the alternative metric, the entire analysis has been repeated and is presented in a (report) and presentation chart (ppt, pdf, movie1, movie2, movie3) formats. The associated latex files, scripts, and data can be accessed here.