Experimental Design Optimization
(This feature is only available in GenEx Enterprise)
Theory
When performing qPCR analysis, the samples are typically prepared in a number of steps before it is analyzed with qPCR, e.g. sample extraction from several animals followed by reverse transcription (RT). Each preparation step introduces variance to the final qPCR data which can be reduced by performing replicates in each step. One could have 3 subjects and take 3 samples for each subject. Each sample is used in 3 RT reactions, and finally each RT tube is split into 3 qPCR tubes. This makes up a 3x3x3x3 nested design with all and all 81 qPCR reactions (see figure).
Under a budget, the number of replicates in each step is limited. If the variation contribution of each step is know, either from a smaller previous trial or approximated, an optimal experimental design can be calculated that minimizes the variance introduced by technical replicates under a given budget.
How to
Enter the data in the Data editor. It must have three classification columns specifying sample replicates, RT replicates, and qPCR replicates, and include one or several different genes. The classification number must be unique for a set of replicates in each column, which means that you cannot have several "nr 1" sets of qPCR replicates. Also, the data must be balanced, which means that there must be the same number of sample replicates for each subject, the same number of RT replicates for each sample, and the same number of qPCR replicates for each RT tube. To estimate the variance contribution in a step, there must be at least two replicates in that level. There must therefore be at least two samples for each subject, at least two RT replicates for each sample, at least two qPCR replicates for each RT. Though, it is allowed to have only one subject, which means that the minimum number of measurements that is allowed is 1×2×2×2 = 8.
To optimize your experimental design, press Experimental Design Optimization button in the Exp. Design tab in to top of the main window.
This will open the analysis in the Control panel where you choose the genes that you want to analyze. There are radio buttons at the bottom of the Control panel that let you either Use data for variance estimation or Input variance contribution by yourself. For the first alternative choose which of the classification columns that represent the different sample, RT and qPCR replicates in the three drop-down lists. It is possible to have two levels, i.e. only one subject. It is important that, if possible, each gene is analyzed with e.g. a t-test or ANOVA, to see whether there is a significant effect to study. Even a good experimental plan cannot find effects that are nonexistent. For the second alternative, input the variance contributions of each level in the three text boxes.
When the genes and classification columns are selected, press the Experimental Cost button down on the left in the Control panel. This is where the cost for each experiment is entered, e.g. the total cost of one RT tube including material, reagents, etc. Enter the total budget for one gene in the Max Cost text field. You can also specify the maximum number of replicates that are allowed for each factor, or leave them at their default values.
To see the results, press the Run button down at the right. The proposed experimental designs are shown in a table together with the total variance and cost for each gene given their designs. The proposed design is the one that minimizes the total variance for a gene but has a total cost under the specified budget. This means that different design can be proposed for different genes.
Warning: Do not use 0 (zero) in the classification columns that defines the groups.