Experimental Design: Optimizing Protocols

Hello again! Today we will be covering our experimental design, and how we are using statistics to save us time and energy.

To recap, the Pitt iGEM Team is developing a genetic engineering toolkitfor P. acnes, because current research on P. acnes is lacking such tools. Onecommon procedure for genetically engineering bacteria is electroporation, where cells areshocked with electricity to create holes (or “pores”) in their membranes.  Once a cell’s membrane is full of pores, DNAsitting outside the bacterial cell will passively slip inside the cell, and the bacteriawill start translating the new DNA as if the DNA were its own. Many differentparameters are involved in electroporation, and different strains of bacteriarequire different conditions for successful electroporation. In our case, thecell wall of P. acnes is relatively strong, so cell wall weakening agents arerequired to help create larger holes when P. acnes is shocked duringelectroporation. Currently, there is no good protocol for inserting DNA into P.acnes with electroporation. Therefore, thePitt iGEM Team is optimizing an electroporation protocol specifically forP. acnes.

At the beginning of May, we developed a list of over 30 variablespotentially influencing electroporation efficiency. We can reasonably do 4 trials a week, so we would need roughly 20 weeks to test 30 differentfactors at three different levels (90 trials / 4 trials/wk), which is farbeyond the scope of a summer project. Therefore, the Pitt iGEM Team is usingthe power of statistics to minimize the number of trials needed to optimizeelectroporation efficiency. Statisticians have developed techniques for identifying themost relevant variables when optimizing a process, known as statistical design of experiments (DOX). DOX is centered around the80-20 rule, where 80% of the optimization comes from changing only 20% of thevariables, so time and money can be saved from only focusing on the few, mostsignificant variables.

One of the pillars of DOX is a method called “partial factor screening,”where only a fraction of the originally estimated experiments need to be run. Ourinitial 90 trials were reduced down to only 16 trials by using a partial factorscreen. Instead of testing 30 variables at three different levels, andmeasuring the effect of variables when changed one at a time, we are testingonly 8 different variables at two different levels, and measuring the effectsof changing multiple variables at once. The power of statistics is provided bythe ANOVA test, where significant variables can be found, even though multiplevariables are changing at once.

The table below listsall our trials and the respective levels for each of the eight variables. Weselected an eight variable partial factor design to maximize the number ofvariables tested, while only conducting 16 trials. The “+” indicates a high level, and a “-” indicates a low level. Seelegend for corresponding values.

 

 

Stay tuned for our experimental results!