## Goal

To provide iGEM teams as well as other scientists a cost and time efficient way to construct genetic constructs.

## Background

One of the overwhelming problems that synthetic biologists face on a daily basis is the variability in genetic constructs. With our ever persistent drive to find a way to develop a standard set of parts for synthetic biology, it is often overlooked that these “standard” parts do not interact the same way that a nut and a bolt do. Sure the threads of each must match for compatibility, but you are guaranteed that a matching set of nuts and bolts will hold together a table just as well as they will hold together a chair, or a piece of metal machinery, or the engine of your car.

When we place together a promoter with a new gene of interest we do not yet have these compatibility rules to know how a promoter will work with this gene of interest. We, as synthetic biologists, still don’t know that one promoter/gene will work the same way in one strain of *E. coli* as they will in another *E. coli*. For synthetic biology to truly emerge as a field that can capitalize in industry, we must find novel ways of engineering biological systems reliably using our own standard parts.

Taguchi method is a statistical method of design of experiment (DOE) that allows information to be obtained from a small number of experiments (1). This method looks at the robustness of a system: how is a system dependent or affected by one part within this system. This is called the Robust Technology Development (2). It is primarily used in industry such as mayonnaise production. For this example a pump, tank, and pipe may be needed to produce mayonnaise of a specific texture and viscosity. These parameters need to be maintained from mayonnaise plant to mayonnaise plant. This is difficult if one plant is in Florida, another in Colorado, and finally in New Jersey. Different plant locations and operating procedures can affect the quality of the product. The goal of the Taguchi Method is to optimize a process based on the mean and variance of these different parameters so that the quality is maintained.

This process was determined by Dr. Geinichi Taguchi, a Japanese researcher whose goal was to create a tool to design an efficient experimental design (2). The goal is to minimize the number experiments; therefore the cost, while still gathering significant data. The following is the philosophy of the method (3):

- Quality should be designed into a product not inspected into it
- Quality is best achieved by minimizing the deviation from a target. The product should be designed so that it is immune to uncontrollable environmental factors
- The cost of quality should be measured as a function of deviation from the standard and the losses should be measured system wide

The method is based on orthogonal arrays which are used to determine the experiments that are to be performed. The arrays are determined by the number of parameters which are referred to as variables and the number of levels of each variable, or states. Each parameter is tested in pairs (3). Once the array is set up, the data from each experiment is used to determine the optimal system setup with minimal variance between systems. Data can be analyzed by ANOVA, bin yield, and Fischer’s exact test or Chi-Squared test as well as other means of statistical analysis (3). In the case of our project, we are looking at how to determine the best system for a genetic circuit. GFP is the gene of interest that is analyzed for ease of gathering and analyzing the data.

## Protocol for Using the Taguchi Method

The following is the typically procedure that is followed for the Taguchi Method:

- Define the process objective:
- The goal is to find a system that has reliable GFP expression regardless of outside noise.
- Determine the design parameters including what affects the performance measurements, such as temperature, and specify the number of levels the parameters will be varied at.
- Parameters are promoters, RBS, and terminators.
- There will be three levels of each parameter.
- Each of the genetic circuits that are constructed by these parameters will be tested in three different strains of
*E. coli*. - Create orthogonal arrays

- Conduct the experiments
- Details on this can be found in the Experimental Design section.
- Complete data analysis

Once the data has been collected, the mean and variance can be calculated by the following equations:

Mean Equation (3)

Variance Equation (3)

Using these two equations, the noise to signal ratio has to be determined:

Noise to signal ratio equation (3)

## Experimental Design

### Experiment

Design experiment to test results of the 27 different circuits in the 3 different strains of *E. coli* for robustness.

### Purpose

The final goal of the project is to determine if the Taguchi Method will give the same results as a full factorial. In this case we want the Taguchi Method to help us determine the most robust circuit design for the expression of GFP. To do this we need to know what the output intensities of GFP are for each circuit in each strain. These can then be analyzed statistically to determine which has the least variance between strains. We are not necessarily looking for the most expression (though one qualifying factor for success is that the GFP is expressed) rather how consistent the expression is between strains.### Procedure

- Standardized Curve Relating OD to Cell Number
- Grow up cells from each strain of
*E. coli*in LB media along with any required antibiotic resistances - Dilute stock cell solution down in 5 ML of LB media so that the initial OD reading is 0.0
- Grow cells at 37⁰C for 6 – 8 hours checking OD every hour to half hour
- Measure a sample of cell solution at each time point with Coulture Counter
- Plot cell number vs. OD for each strain to obtain standardized curve
- 3A Assembly of Constructs Expressing GFP
- From the Taguchi Orthogonal Array, determine which genetic circuits to construct
- Using iGEM’s protocol for 3A assembly
- Calculate the GFP Expression/Individual Cell
- After all three strains are transformed with each construct for a total of 81 constructs to test, use Cary Eclipse Fluorescence Spectrophotometer and OD to gain total GFP expression and number of cells
- Divide total expression by number of cells to gain GFP expression per cell
- Statistically Determine Which Construct has Most Robustness
- After the amount of GFP expression per cell is calculated for all 81 constructs, use JMP statistical Analysis Software to analyze the data.
- Using the screening or fit model platform and the built in optimizer in JMP, see which construct is the most robust.
- Compare the robustness for the 81 constructs against the Taguchi constructs to see if the most robust construct in the 81 constructs array is a part of the Taguchi array.

## JMP Statistical Analysis of Results

The mathematical representation of the RSM model for first order linear equation and second order quadratic equation are as follows:

__First order linear equation model with interaction__

__Second order quadratic equation model __

- Choose Fit Model command in the Analyze menu
- Click the variables to ender model effects
- Select one or more columns in the column selector list
- Choose
**Response Surface**from the**Macros**pop-up menu - Choose
**Screening Fit**from the fitting pop-up menu and click**Run Model** - You can select Sum of Squares of the difference to be minimum by selecting Standard Least Squares to solve for the critical values.
- After fitting the data to the RSM, the mixture profiler can be used to visualize and optimize response surface resulting from the multiple variables.

__Here is a JMP example__

This example comes from an experiment aiming to optimize the texture of fish patties. The columns Mullet, Sheepshead, and Croaker represent the proportion of the patty that came from that specific fish type. The column Temperature represents the oven temperature used to bake the patties. The column Rating is the response and is a measure of texture acceptability. A response surface model was fit to the data and the prediction formula was stored in the column Predicted Rating.
__Here is a JMP example__

To launch the

**Mixture Profiler**, select

**Graph**>

**Mixture Profiler**. Assign Predicted Rating to

**Y**,

**Prediction Formula**and click

**OK**.

#### The initial profiler should look something like this:

__Using this procedure to find the most robust genetic construct__

For our project, the variables or the input factors were the promoters, Ribosome binding sites, terminators and the strains of bacteria. The response is the GFP expression per cell.
Using JMP, and the results from the experiment, you can analyze what factors are significant and what combination of these factors will results in the most robust genetic construct to give optimized protein expression per cell.
__Using this procedure to find the most robust genetic construct__

## Procedure for Designing a Screening Experiment in JMP

- Go to
**DOE**>**Screening Design** - Add the number of variables to either of the option depending on the experimental design.
- In the case of our project, we have four factors and each of the four factors has three categories. So we add four 3-Level Categorical. This example shows you a 3, 3-Level Categorical.
- Fill up the factors and the three categories
- Choose the design type you want. You can have either
**Taguchi**or**Full Factorial** - Leave the number of center points and replicates as 0 because they increase the number of runs. Click Make Table.

## Collected Data

### Taguchi Arrays

When choosing the Taguchi Array, there pre-set arrays based on the level of noise. For our experimental design we choose L9 Array:

### Growth Curve

Growth curves of the three different species of *E. coli* were taken. The purpose of this is to set up the standard curves. The next step is to use a Coulter Counter to determine the number of cells at each OD. This would give the standard curve for determining the amount of GFP expressed per cell.

## Challenges and Failures

### Transformations

There were many different factors that contributed to unsuccessful transformations. The first was improper lab technique as we went through the summer learning the basics of aseptic technique and growing cultures. The next was cells that had been thawed too many times rendering them incompetent which was not realized until later on in the process. Finally, changes in our working space altered the growing conditions so as to not be optimized for*E. coli*growth which set us back in our schedule.

We learned that although we take for granted the beneficial act of bacterial cells up-taking DNA, these cells cannot maintain a state of competency forever. In a natural environment this would increase their vulnerability to outside toxins that could also enter the cells through engorged pores.

### 3A Assembly

Our biggest mistake in the process of 3A assembly was purifying the DNA after digestion and ligation. Due to lack of lab equipment, purification via gel extracting was unavailable until just recently. Ethanol purification was tried, but failed as it seemed most of the DNA became trapped in the supernatant. This may be due to not enough sodium acetate for the DNA to precipitate out. Nonetheless, our DNA results were low which resulted in continued failed attempts at transformations.## Future Work and Changes

1. Finding promoter+RBS parts to reduce the amount of 3A assembly as well as the number of tests that need to be done.

2. Gaining the use of a gel purification kit for increasing our DNA results.

3. Finding a lab space with correct

*E. coli*growth conditions that can be closely monitored.

## References

- Venil, C. K., & Lakshmanapermalsamy, P. (2009). Taguchi experimental design for medium optimization for enhanced protease production by bacillus subtilis hb04.
*Journal of Science & Technology* - Rao, R. S., Kumar, C. G., Prakasham, R. S., & Hobbs, P. J. (2008). The taguchi methodology as a statistical too for biotechnological applications: A critical appraisal.
*Biotechnology Jornal*, (3), 510-523. doi: www.biotechnology-journal.com - Fraley, S., Oom, M., Terrien, B., & Zalewski, J. (2007, November 27). Design of experiments via taguchi methods: Orthogonal arrays. Retrieved from https://controls.engin.umich.edu/wiki/index.php/Design_of_experiments_via_taguchi_methods:_orthogonal_arrays
*JMP: Statistical Discovery from sas.*(n.d.). Retrieved from http://www.jmp.com/- Alexander, M. (n.d.).
*Response Surface Optimization Using JMP Software.*