Sample Variation
Most importantly one should consider how to minimize the amount of sample variation. If at all possible, control the research subjects for gender, age, and genetic background to minimize the experimental variation. A simple experimental design, where the subjects are only different based on the research question of interest, will yield robust microarray results. The following variables could influence mRNA expression and should be controlled:
- Gender
- Age
- Litter
- Genetic background
- Other environmental or treatment differences, i.e. disease, drug treatment, radiation exposure
- Laboratory Protocols, i.e. treatment, RNA extraction. It is critical to maintain consistency.
If it is not possible to absolutely control your research subjects, try to control as much as possible. For the variables that cannot be controlled, keep them as balanced as possible amongst the differentc onditions/pheonotypes/treatments of interest. E.g. For wild type samples and treated samples, if the gender can't be controlled, try to collect samples in a balanced way, so that the uncontrollable variable will be averaged out and will not interfere with your research interest.
Sample ordering, array platform, and processing method must be taken into consideration. Please follow the guidelines below:
- Consider batch variation, i.e. processing day of samples, day/time harvesting cells. We strongly recommend that you process all samples together or in batches where the phenotypes of interest are equally mixed. Do not process all wildtype samples on one day and all treated samples on another. When submitting samples to our facility, please keep in mind that we process samples in batches of 12, so please order your samples on the submission form such that the phenotypes are distributed evenly into several batches (depending on the number of samples submitted).
- Maintain the same array platform. Always use the same array type for your experiment, i.e. do not use Human Genome U133A for some samples and Human Genome U133 Plus 2.0 for others.
- Maintain the same array processing methods, i.e. Affymetrix 3' IVT is not the same as NuGEN Amplification or Ambion MessageAmp processes. Amplified and nonamplified data cannot be compared.