|
|
|||
|
|
||
|
|||
|
Dana Campbell, Lisa Fan, Keith Roby and Graham Threadgill, Beckman Coulter, Inc.; and Pat Whitcomb, Stat-Ease*, Inc. IntroductionOptimizing biological assays conditions is often a challenging process facing scientists. The demand to produce quality and robust assays that work across a range of biological conditions is often strived for along with a short development timeframe. In addition, automated systems are often required to enable scientists to screen in a high-throughput environment. In principle, this goal has a clear objective. However, setting up an experiment that systematically studies all experimental parameters is a challenge. Often times, the one factor at a time (OFAT) approach is used to systematically study individual experimental conditions. While this approach uncovers how changing a single factor will affect assay conditions, it does not uncover interactions that may exist between experimental factors, thus leaving out critical information that may significantly reduce assay quality and robustness. In light of this, scientists have been adopting design of experiment (DOE) methodology using mathematical models to evaluate experimental parameters to optimize assay conditions. Designs are created that evaluate both main effects and interactions between experimental factors such as sample concentration, reagent types, and incubation time. These experimental parameters represent only a few of the many influencing assay factors that can be studied using DOE. Statistical software packages are readily available that create experimental designs. However, translating statistical designs to run on automated liquid handling systems is a complex endeavor. In order to address this, Beckman Coulter, Inc. has developed a software package Automated Assay Optimization (AAO) FX Software for the Biomek FX liquid handling system. AAO FX software (AAO FX) allows scientists to import designs that are translated into corresponding Biomek FX methods. Design models created in statistical software packages such as Design-Expert* (DX6) software developed by Stat-Ease, Inc. (Minneapolis, MN) offers scientists a straightforward approach in creating an array of design models such as two level factorials, D-optimal designs, split-plot design, etc. AAO FX guides the user through a wizard-like interface to specify experimental parameters such as labware, deck configurations and pipetting options to generate Biomek FX methods. Experimental data can be deconvoluted back into the original design format and transferred into a statistical package for analysis. Design-Expert software provides in-depth analysis with interactive graphs to assist in interpreting assay results. Integrating DOE and automated liquid handling technology offers scientists the ability to design experiments that test a wider dynamic range of assay conditions. This allows for a clear understanding of the assay and its components so that improvements can be implemented and the assay optimized. In this application study, we demonstrate how to utilize these powerful software packages to optimize a multiplex PCR to be run on the CEQ™ 8000 Genetic Analysis System (CEQ). Combining multiple primers into a single PCR is difficult, as many factors and interactions will influence the PCR* results. However, once the conditions are optimized and set for a given group of primers, scientists are rewarded with large reductions in valuable laboratory time and resources. In this study, we demonstrate how DOE and AAO FX is used to optimize a multiplex PCR using three primer pairs. The primer pairs were obtained from ResGen Invitrogen* Corporation, Inc. (Carlsbad, CA). The primers used in this study are part of the Weber panel of primers used for genotyping human DNA. The regions of DNA amplified by these primers are short tandem repeat regions (STR's) and the primers are often referred to as STR primers. The multiplex PCR should produce 3 distinctly sized markers representing the individual's specific genotype. By using the powerful combination of experimental design and analysis software with automation offered by Beckman Coulter, Inc., we demonstrate how complex assay optimization can be simplified. The process of using DOE and AAO FX for assay optimization is illustrated in Figure 1.
Materials and Methods:
For this experiment, five experimental factors and a goal to optimize a multiplex PCR has been defined. Using Design-Expert software, a split plot design was created using the following five factors at two levels: three primer set concentrations each labeled with a different fluorescent dye, MgCl2 concentration, and annealing temperature. The annealing temperature factor was associated with the thermocycler while the remaining four factors were associated with the Biomek FX. Using the split plot design restricted the randomization based on the annealing temperature factor. One plate contained the annealing temperature factor at the low level while the other plate contained the annealing temperature factor at the high level. This split plot design used all combinations of the five factors; therefore, 32 separate experimental runs were possible (see Figure 2). To increase the probability of finding a significant effect in this experiment, we decided to run a replication. The probability of finding an effect for a model increased as the ratio of the effect to noise increased. Replicating this experiment to 64 experimental runs increased the probability of detecting main effects to >97% at 1 standard deviation.
Figure 2. A subset of the first experimental design layout created with DX6 showing the first 20 of 64 experimental runs in both standard and run order. Note the design is kept in standard order since AAO FX software will randomize the runs with an experimental plate when it creates the Biomek FX method. The column headings show the 5 factors being studied and their respective final PCR concentrations or temperature unit. The low and high concentrations for the primers are 0.015pM and 0.08pM respectively. The low and high concentrations for the MgCl2 are 1.5mM and 3.0mM respectively. The low and high annealing temperatures are 50 °C and 60 °C respectively.
The design parameters from the import file can be specified in AAO FX along with experimental plate formats and well restrictions if necessary. AAO FX supports both 96- and 384-well formats. AAO FX offers the opportunity to select randomization options if more than one plate is required for the experiment. This can also be established in the design file itself through blocking. Once these considerations have been determined, AAO FX will randomize the design into the selected experimental plate format to generate a plate map. The plate map is a graphical representation of the experimental plates containing color-keyed identifiers illustrating the content of each randomized well on the plate (see Figure 3).
Results:We first ran four replicate multiplex reactions using our standard PCR conditions, giving us a baseline as to how well the primer pairs work together. The following were the default factor conditions used in the experimental design:
The electropherogram from this experiment is shown in Figure 6. In this run, the blue marker (D4) gave a low signal, which was called correctly by the CEQ software, but could clearly be optimized. Overall, the signal balance from the three markers was inconsistent and without optimization this could lead to misinterpretation of the data. A blue peak in the 131 nt range appeared throughout the runs, even after optimization. This artifact was not in the expected STR range for the D4 marker and therefore did not interfere with the analysis. The first experimental run consisted of five factors at two levels each. We used three different STR primers in which heterozygosity was highly probable. We expected to see two main peaks for each marker representing each allele. The forward primers beach had a different colored florescent dye that labeled them either blue, green, or black (D4, D3, D2 respectively). Each primer pair concentration was a factor to be studied.
The split plot design tested multiple variations of different factors for a total of 32 unique runs (experimental reaction setups). In order to measure the response of the reaction to the different conditions, the ratio of the higher of the two alleles peak heights to the peak height of the standard at 200 nt was measured for each primer pair. Figure 7 illustrates a CEQ result from our first experimental run. This particular sample has all three heterozygous markers represented. As predicted, some of the conditions had detrimental effects while others had improved effects.
Design-Expert software was then used to analyze which effects were significant and determine the success of the experiment based on the data collected (see Figures 8, 9 & 10).
Using the model graphs in Design-Expert software, it was clear that MgCl2 concentration had a large impact on the data obtained for the D4 marker. All three response model graphs were analyzed and it was determined that concentration of the primer pairs interacting with the MgCl2 concentration, were significant factors influencing how well the reactions would work. Figure 8 is part of the analysis section of Design-Expert software that generates a half-normal plot of effects to help determine which effects are significant. Figure 9 confirms the statistical significance of these factors. Figure 10 displays the diagnostic graph that statistically validates the model created showing factor effects. We took advantage of the optimization module within Design-Expert software. This searches for a combination of factor levels that simultaneously satisfy the requirements placed on each of the responses and factors. Optimization of one response or the simultaneous optimization of multiple responses can be performed graphically or numerically. Assay optimization is an iterative process and although the multiplexed PCR was now giving far superior data to the original conditions, further optimization was possible, particularly in the area of balancing the signals from the three different markers. Data from the first optimization indicated that annealing temperature did not have a significant impact within the range tested (50 °C and 60 °C). Therefore, in the next design, that factor was omitted and two different levels of the remaining factors were studied. This 24 full factorial design created with four replicates used all combinations of the four factors generating 64 experimental runs. This increased the probability of detecting two main effects and a two-factor interaction to 97.6% at 1 standard deviation. All reaction volumes were the same as for the first design. The low and high concentrations for primers and MgCl2 were altered based on the data from the first round of optimization. Data from those experiments indicated that the concentration of D2 should be higher than D3 and D4. For this round of optimization, the low and high concentrations for D2 were 0.05pM and 1.0pM; for D3, 0.04pM and 0.09pM; and for D4, 0.03pM and 0.06pM. The low and high concentrations for MgCl2 were now 1.0mM and 2.0mM as opposed to 1.5mM and 3.0mM in the first series. As can be seen in Figure 11, the series of experiments led to further optimization of the assay.
The second series of experiments also revealed some interesting two-factor interactions. The best data was obtained when D2 was at its higher concentration (1.0pM), D3 at its lower concentration (0.04pM), D4 at its higher or lower concentration (0.03pM or 0.06pM) and MgCl2 at its higher concentration of 2.0mM. It was also observed that increasing the concentrations of all primers could have a detrimental effect. Another interesting interaction that came to light was that marker D2 did not work well when D3 was at its higher experimental concentration. Comparing the electropherogram from Figure 11 to Figures 6 and 7, it is obvious that two rounds of optimization significantly improved this multiplexed PCR. Summary:We have demonstrated that using statistical experimental design and Biomek FX AAO software together quickly and efficiently optimized this complex reaction. The experimental design and automation described in this application can be extrapolated to a variety of optimization and screening applications. The powerful combination of design and analysis software with robotic method development software for the Biomek FX greatly enhances the ability of scientists to meet the challenge of assay optimization. Assay quality was significantly increased without sacrificing valuable resources and precious samples. This process clearly demonstrates how DOE can be employed with automation allowing complex designs to be executed in a straightforward and efficient manner. * All trademarks are the property of their respective owners. Where applicable, the PCR process is covered by patents owned by Roche Molecular Systems, Inc., and F. Hoffmann-LaRoche, Ltd.
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
There are several important assay considerations that must be
evaluated prior to creating an experimental design. First, the user
must have a clear understanding of the experimental protocol so
that significant factors are chosen with appropriate dynamic ranges
to study. Automation considerations such as available resources,
labware types, and calibrating the automated system for accuracy
and precision is critical to ensure that additional noise is not
introduced into the experimental model. Lastly, a clear objective
must be defined along with a good understanding of how the data will
be analyzed.
Design-Expert software can export the design file into an AAO FX
readable text file. AAO FX accepts any design in text or comma-delimited
(*.csv) files but may require modification to adjust the files into the
appropriate AAO FX format.
Once the plate map has been generated, AAO FX guides the user
to create the liquid handling method. During this stage, the user
has control over the deck allocation, order of pipetting, and
associated parameters such as specifying volumes, aspiration and
dispensing speeds, heights, mixing, and tip touches. Pipetting
settings can be modified and saved with user-assigned names for
use in future experiments, producing significant time savings.
Assay components that are not factors in the experimental design,
such as reagents or background wells can be added and configured
independently from the experimental design. Critical time sections
can be added to a method where the Biomek FX will configure an
individual time-scheduling protocol for each well. Once method
construction is complete, AAO FX automatically validates it. The
generated Biomek FX method is illustrated below in Figure 4.
Once the liquid-handling method is prepared and validated, the
assay is ready to be run. The user is responsible for loading all
labware and associated reagents onto the Biomek FX deck (as
graphically guided by the software). AAO FX generates an HTML
document that provides a report for the required labware and
experimental plate. As seen in Figure 4, the method consisted
of nine reagent transfers. The first transfer consisted of the
Span-8 aspirating 9 uL of PCR master mix from a single source
and dispensing to all 64 experimental wells. The following eight
transfers consisted of the Span-8 aspirating 4 uL of specific
concentrated reagent and dispensing into 32 of the 64 wells. The
subsequent transfer steps pipette the specified reagents to the
randomized experiment wells based upon the generated plate maps
illustrated in Figure 3. The two experimental plates each
contained 32 experimental wells to be thermocycled at the two
specified temperatures and run on the CEQ.
Upon completion of the assay, experimental data can be collected
using the AAO FX Data Collection Utility. This utility is used to
copy or import result data from analytical equipment such as a plate
reader. A Microsoft* Excel workbook is created illustrating the
experimental plate map that contains experimental well information
(see Figure 5). In this experiment, the results were analyzed
manually by ratio analysis of peak heights (experimental: size
standard) for each PCR product. After the transformation of the
data, it was entered into the generated worksheet for each plate.
Then, AAO FX deconvoluted the data into the original design format
to be pasted into a report worksheet.
Next the data is transferred into a statistical software package
such as Design-Expert software for analysis.
