Contents - Index

Pre-processing of Data

The Pre-processing menu in the Data editor have several procedures that can be applied to the data before the acctual analysis.

The proper order of the pre-processing operations depends on the experimental design! It is important to make sure that the pre-processing steps are done in the correct order, since they do not always produce the same result if the order is different. Most operations can be performed in opposite order, but not all of them. With nested experimental designs e.g., the technical repeats should be normalized starting from the lowest level (typically qPCR repeats), going upwars to the highetst level (typically subjects). Data that will be analyzed in linear scale must be converted to relative quantities before averaging technical repeats. For data that will be analyzed in logarithmic scale, the order of averaging technical repeats and normalizing with reference genes does not matter. qPCR repeats should be averaged before normalization with reference genes, while RT repeats should be averaged after normalization with reference genes. The guide below shows a suggested order of pre-processing operations, but it might not be suitable for your specific experimental design.

These are all the pre-processing operation that GenEx offers. You are not supposed to use all of them on the same data set, since the use of some operations makes the use of others redundant. 

• Cut off: Lets you define a cut-off Cq value to handle off-scale data, such as primer-dimer signals.
• Missing data: Handles missing data by filling them with specified values. 
• Test for outliers: Opens a window to perform Grubbs' test for detecting outliers in the data.
• Normalization with Interplate Calibrators: Compensates for differences between separate runs by normalizing with interplate calibrators.
• Efficiency correction: Corrects for PCR efficiencies. It is possible to define different PCR efficiencies for different columns (genes). 
• Normalize with Spike: Correct for difference by normalization to a spike.   
• Normalization to sample amount: Compensates for differences in sample amounts.
• Average qPCR repeats: Averages qPCR repeats based on a classification column defining which of the samples that belong together.  
• Normalize to reference gene(-s): Lets you normalize with reference genes.
• Average technical repeats: Averages technical repeats e.g. RT or sampling replicates. Does the same thing as Average qPCR repeats.
• Normalize with reference sample(-s): Normalize with reference samples.
• Relative quantities: Calculates relative quantities. Compares the quantities in all samples against the quantity in one certain sample.

Some statistical analyses assume that the data is normally distributed, and qPCR data tends to be normally distributed more often on logarithmic scale than on linear scale. Therefore, GenEx offers four options of converting the data to logarithmic scale. 

• Log 2: Converts all values to log base 2 (fold changes). New value = log2(Old value).
• Ln: Convert all values using the natural logarithm (base e). New value = ln(Old value).
• Log 10: Convert all values to log base 10. New value = log10(Old value).
• Log(1 + X): Converts all values +1 to log base 10, which makes it possible to convert to logarithmic scale even if there are measurements of zero. New value = log10(1+Old value).

Scaling can be used to easily visualize data of different magnitudes, or to get better results from classification methods since a lot of methods depend on the magnitudes of the measurements. 

• Mean center: Mean centers the data so that the mean of each column (gene) is zero. 
• Autoscale: Autoscales the data so that the mean and standard deviation of each column (gene) is zero and one respectively. 

If you have any questions, don't hesitate to contact MultiD for support (support@multid.se)!