EXPERIMENTAL DESIGN AND ANALYSIS FOR MICROBIOLOGY: COURSE SUMMARY

ERICK TATRO Abstract. This is a summary of a laboratory course proposed for advanced microbiology students. It assumes basics skills that would have been acquired in a comprehensive microbiology curriculum. The focus is on statistical analysis of data, modeling, prediction, and experimental design. The goal is to prepare students for work in elds like biotechnology, pharmaceutical manufacturing, and research & development. It is also a writing intensive course, wherein the studies are expected to prepare detailed pre-laboratory reports and prepare formal reports. A They will prepare the reports by lling out L TEXtemplates, which will progressively contain less structure until by the end of the course they are able to create their own scientic document. The course uses problem-based learning, involves group and solo work, and emphasizes professionalism. The students should be excited about their nal solutions and be encouraged to envision potential uses in other elds for the skills gained in this course. The full working document of the course manual can be accessed at the following url: https://docs.google.com/file/d/0B5DOUvfVzRAITDh2T3l5bXYxSW8/edit?usp=sharing

1. Learning Goals (1) Understand experimental variance, eciency, and balanced experimental design. (2) Use JMP statistical software Design of Experiments platform to plan and model experiments and for hypothesis testing. (3) Graphically represent data and results. (4) Design factorial experiments with interaction terms. (5) Augment experiments by incorporating additional runs and analyzing appropriately. A (6) Prepare scientic reports and documents using L TEX. (7) Work as a team. (8) Understand that I-optimal design criterion focuses on precise parameter predictions, for response surface designs. (9) Understand that D-optimal design criterion focuses on estimation, for screening experiments. (10) Design experiments involving mixtures. (11) Design experiments using co-variates, known factors that vary, aect the outcome, but the experimenter cannot control.

Date : September 18, 2013.

1

Microbio Course Summary, Erick Tatro

2. Grading Final grades will be calculated according to the following: Pre-lab Report . . . . . . . . . . . 20% Post-lab Grade . . . . . . . . . . 30% Lab Reports . . . . . . . . . . . . . 50% 3. Materials 3.1. Text. (1) Peter Goos and Bradley Jones. 2001 Optimal Design of Experiments: A Case Study Approach. John Wiley Sons.1 (2) SAS Institute Inc. 2012. JMP c 9 Design of Experiments Guide. Cary, NC: SAS Institute, Inc.2 3.2. Required Software. The course is taught toward using Versions 9 or 10 of JMP software. A L TEXdocument preparation software is required (2001 version or later).3 4. Prerequisites Microbiology laboratory course, Mathematics to Calc I, writing course. 5. Schedule Outline Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Week 1 M Experiment 1, Optimal Design With Dierent Variances . . . . . . . . . . . . . . . . . . . . . . . . . Week 1 W Experiment 1, Optimal Design With Dierent Variances . . . . . . . . . . . . . . . . . . . . . . . . . Week 2 M Experiment 1, Optimal Design With Dierent Variances . . . . . . . . . . . . . . . . . . . . . . . . . Week 2 W Experiment 2, Screening Experiment, Optimizing Transfection Eciency . . . . . . . . . Week 3 M Experiment 2, Screening Experiment, Optimizing Transfection Eciency . . . . . . . . . Week 3 W Experiment 2, Screening Experiment, Optimizing Transfection Eciency . . . . . . . . . Week 4 M Experiment 3, Surface Response with Categorical Variable, Optimal Medium . . . . . Week 4 W Experiment 3, Surface Response with Categorical Variable, Optimal Medium . . . . . Week 5 M Experiment 3, Surface Response with Categorical Variable, Optimal Medium . . . . . Week 5 W Experiment 4, Mixture Design, Optimal Antibiotic Formulation . . . . . . . . . . . . . . . . . . Week 6 M Experiment 4, Mixture Design, Optimal Antibiotic Formulation . . . . . . . . . . . . . . . . . . Week 6 W Experiment 4 Mixture Design, Optimal Antibiotic Formulation . . . . . . . . . . . . . . . . . . . . Week 7 M Experiment 5, Irregular Design Region, Optimal Incubation . . . . . . . . . . . . . . . . . . . . . . Week 7 W Experiment 5, Irregular Design Region, Optimal Incubation . . . . . . . . . . . . . . . . . . . . . . Week 8 M Experiment 6, Targeted Outcome, Fermentation Experiment . . . . . . . . . . . . . . . . . . . . . Week 8 W Experiment 6, Targeted Outcome, Fermentation Experiment . . . . . . . . . . . . . . . . . . . . . Week 9 M Experiment 7, Prediction Modeling, Food Microbiology . . . . . . . . . . . . . . . . . . . . . . . . . . Week 9 W
1Electronic version available from the University library 2PDF available from www.jmp.com/support/downloads/pdf/jmp1001 3Open source software and documentation at www.latex-project.org

Microbio Course Summary, Erick Tatro

Experiment 7, Prediction Modeling, Food Microbiology . . . . . . . . . . . . . . . . . . . . . . . . . Week 10 M Experiment 7, Prediction Modeling, Food Microbiology . . . . . . . . . . . . . . . . . . . . . . . . Week 10 W Finish Experiment 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Week 11 M Make Up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Week 11 W Make Up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Week 12 M 6. Introduction to Experiments 6.1. Experiment 1: Optimal Design With Dierent Variances. 6.1.1. Reading Assignment. (1) Chapter 1, SAS Institute Carry out the exercise. (2) Chapter 1, Goos and Jones 6.1.2. Your company has tasked you with determining the average output from two dierent bioreactors that should be producing the same thing. The bioreactors contain bacteria and the machines use a nozzle to dispense 100 L. One of the bioreactors is several years older than the other and it uses an older design of nozzle. It therefore dispenses 100 L with less precision than the other machine, even though it is producing the same amount of bacteria. As part of weekly Quality Assurance, you need to measure the colony forming units (cfu) per mL in the bioreactors each week and assure that they are producing the same amount. There are enough resources to perform ten total measurements on one day each week. You are budgeted with 12 runs to experiment with methods to decide on the most ecient standard operating procedure for this weekly measurement. 6.2. Experiment 2: Screening Experiment, Optimizing Transfection Eciency. 6.2.1. Reading Assignment. (1) Chapter 2-3, Goos and Jones (2) Chapter 2-3 and 13, SAS Institute 6.2.2. You work in product development at a biotech company that is about to bring a bacterial transfection kit to market. The kit is composed of a proprietary bacteria strain and plasmid with the -Galatosidase marker gene, so successfully transfected bacteria will show a blue colony on an X-Gal agar plate. You need to develop the components of the kit and write the product insert that explains to the end-user how to use the product. You are assigned with determining the following: X1 : Quantity of bacteria in the transfection reaction; X2 : CaCl2 concentration included in the kit; X3 : Plasmid concentration included in the kit; X4 : Time in cold-shock recommended in the instruction manual. Design an experiment to maximize the transfection eciency. This time, your project manager has budgeted for 16 runs in total. Youve decided to do an initial 12-run experiment and then augment with the remaining allotted budget to verify your conclusions. 6.3. Experiment 3: Surface Response with Categorical Variable, Optimal Medium for Maximal Growth.

Microbio Course Summary, Erick Tatro

6.3.1. Reading Assignment. (1) Chapter 4, Goos and Jones (2) Chapter 4 SAS Institute. 6.3.2. You are the production manager at a therapeutics pharmaceutical company that uses bacteria to create recombinant proteins which are then used to treat diseases. You produce 3 dierent therapies which grow in three dierent strains of bacteria. Because they were developed at dierent times, the strains grow in three dierent media formulations. After meeting with your counterparts who manage supply chain and costs, you have decided that maintaining 3 media formulations is inecient. It uses space, labor, and resources to manage the creation, storage, tracking, and purchasing. You have therefore decided to generate a single formulation for all 3 strains which produces a maximal growth rate for all three strains. You have X1 : Strain 1, Strain 2, Strain 3; and you must decide on the amount of the following: X2 : beef extract, X3 :yeast extract, X4 : peptone, and X5 : NaCl. 6.4. Experiment 4: Mixture Design, Optimal Antibiotic Formulation. 6.4.1. Reading Assignment. (1) Chapter 5, Goos and Jones (2) Chapter 7 SAS Institute. 6.4.2. You work for the Center for Disease Control and Prevention. There has been an outbreak of an unusual nosocomial bacteremia at hospitals in a particular region. You have been tasked with deciding what combination of antibiotics to prescribe that all three pathogenic strains will be susceptible to while not killing the mutualistic microbe. You will be given a panel of fourteen antibiotics to work with and will test the pathogens susceptibility using the Kirby-Bauer Disk Diusion method. Each laboratory (student) will be testing dierent combinations of unknown pathogens and a culture of Lactobacillus reuteri. Each of the fourteen antibiotics have dierent costs associated with it and you will be provided with a list of weights assigned to each antibiotic. After testing the antibiotic susceptibility of your pathogens, prepare a report recommending a combination of two to three antibiotics whose total cost does not exceed 100 and confers at least moderate susceptibility to all three pathogens and at most intermediate susceptibility to the mutualistic bacteria. It will be formulated into one pill and sent to the aected region. 6.5. Experiment 5: Irregular Design Region, Optimal Incubation Conditions. 6.5.1. Reading Assignment. (1) Chapter 5, Goos and Jones (2) Chapter s 1 and 7 (page 164-166) SAS Institute.

Microbio Course Summary, Erick Tatro

6.5.2. You are the manufacturing process manager for a therapeutics company that develops and manufactures treatments that are produced as a recombinant proteins from stably transfected E. coli. The company wants to set up a new section of bioreactors for scaling up and manufacturing its latest product that was recently approved for the market. Expression of the therapeutic protein is induced by arabinose through the ara C operon. Inducing expression is facilitated by aeration, so shaking is important. The development team has been working at the 5 mL preparation scale and you need to scale it up to larger volumes. Design an experiment that optimizes the following parameters: X1 :Incubation Temperature, X2 : arabinose Concentration, X3 : Volume, and X3 : Shaking Speed, which yields the highest protein concentration: Y . You cannot allow the reaction mixture (media) to reach 80% of the height of the reaction container (1.5 L asks), therefore as shaking speed increases, the volume must be lowered to accommodate. First, determine the relationship between shaking speed and volume as a constraint for your experimental design. It should t to some equation. C < 1 X3 + 2 X4 + 3 Where C is some constraint and 1 , 2 , 3 are coecients empirically determined in the laboratory. After designing and carrying out your experiment, you will determine the optimal parameters that induces the greatest expression of the protein therapeutic, which you will measure based on its uorescence under UV light. Prepare a report to your manufacturing team recommending the optimal set up. You will work in teams of 4, and you can only test 3 dierent shaking speeds and 3 dierent temperatures, and you are budgeted with a total of 12 runs. 6.6. Experiment 6: Targeted Outcome, Fermentation Experiment. 6.6.1. Reading Assignment. (1) Chapter 6, Goos and Jones (2) Chapter 7 (page 168-178), SAS Institute. 6.6.2. You are an entrepreneur ready to retire early from your years of service in the biotechnology industry. Since you cannot stand the thought of a non-productive retirement, you decide to nally live out your early dream of opening a brewery and are ready to test out your beer and wine making skills. You are taking on the challenge of designing the optimal concentrations of sucrose (X1 ), maltose (X2 ), hops (X3 ), barley (X4 ), and malt (X5 ) (or grapes (X3 ), if making wine). You are able to purchase all the components you need and test out your brews using jugs in the lab. However, you will want the just so alcohol by volume (ABV) for your particular brew (Y ). Design an I-optimal experiment, constructing a ternary plot of the experimental design showing the ve parameters and twelve runs. You will measure the specic gravity using a hydrometer before and after fermentation to calculate the ABV. This is a two week long experiment in three teams, where we will scale down the volume to 500 mL and brew three batches of 12 runs of beer or wine. Each student is responsible for their own pre-laboratory report with experimental design plans and the team will build and implement a consensus plan. The plan will be implemented this week, and two weeks from

Microbio Course Summary, Erick Tatro

now, the ABV calculated. You then will prepare a report determining the optimal proportions of the ingredients to achieve 5.5% ABV for Beer 1, 7% ABV for Beer 2, and 9% ABV for the Wine. 6.7. Experiment 7: Prediction Modeling, Food Microbiology. 6.7.1. Reading Assignment. (1) Chapter 9, Goos and Jones (2) Chapters 1 and 7 (page 164-166) SAS Institute. 6.7.2. You are a sta scientist at the Food and Drug Administration and have been given epidemiological data on growth of Listeria monocytogenes from food containers (e.g., refrigerators), from vendors across the country. You are given conditions on temperature, moisture content, and Listeria growth. Use the DOE platform and outcome data to make a model predicting the risky growth conditions suitable for Listeria. Test your predictions in the laboratory. You will be given several test food items whose moisture content is known, but you cannot control, and will inoculate with Listeria and allow incubate in various low-temperature food containers and then measure Listeria growth. Based on your experimental data and epidemiological analysis, prepare a report recommending regulations requiring maximal temperature for ranges of moisture content. 7. Comments 7.1. Course Structure. The course is designed to learn through personal and group accomplishments. It emphasizes planning, record keeping in the laboratory notebook, professionalism, and reporting conclusions. Even though there are relatively few laboratories, the planning, analysis, and write up take signicant amount of time and will require signicant work outside of the laboratory. Practical applications and role-playing facilitate learning for a diverse group of students and this course is designed to teach real world skills and stimulate thought on applicability. Intensive writing facilitates development of critical thinking skills and the writing portions of the course wherein the students are required to make a recommendation supported by evidence gathered in the laboratory will strengthen critical thinking. 7.2. Alternatives and Additions. I considered using open-source DoE packages in the statistical language, R, which would cost less for the students. However, the lack of a graphical user interface and learning curve for R may pose too high of a challenge and be a distraction when the purpose of the course is to master other skills. Typically, science students (except for A computer science or physics) are not introduced to L TEXdocument preparation until graduate school (if at all). Students will be at an advantage if they learn this skill during advanced undergraduate stage instead of relying on wysiwyg word processing software. It also exposes them to Open Source software and versatile usage of computers. I considered two additional experiments to this course outline pertaining to analyses of microbial evolution. These experiments involved using Clustal for sequence alignment software, constructing cladograms, and placing an assigned unknown sample sequence within a clade, and then experimentally verifying predictions with traditional microbiology techniques. The other experiment involved analyzing

Microbio Course Summary, Erick Tatro

a bottleneck eect of inoculating medium with small samples from stocks and the new medium containing some selection agent (e.g., progressively higher salinity), and then using DNA sequencing to determine if particular allele frequencies had changed. While these experiments do t with the overall theme of data analysis, and statistics, it seemed outside the scope of experimental design. 7.3. Conclusion. I conceived of this laboratory course while discussing Research and Development principles with colleagues in the biotechnology industry, specically in product development. I learned that there is a dearth of knowledge by life scientists in this knowledge area. Graduates in microbiology, or any life science, with statistical knowledge, skills with software platforms, skills with graphical representation of data, and Design of Experiments principles will have an advantage in the job market. These skills are also useful in academic research in planning aims, designing hypotheses, optimizing protocols, and optimally utilizing resources. The role-play model of the course manual is designed to be stimulating, facilitate discussion, and also enable to students to visualize how their education is will be used in the future.