2 replies to this topic

[NeuroGuy]

I have no clue on how to do the statistics for my Western Blot experiment.  The experiment was looking at expression of a protein of interest (as well as a control protein) in 3 genotypes (wild-type, hemizygous, homozygous) at three timepoints (1, 2, and 3 weeks).  For each timepoint, I collected 3 samples of each genotype (= 9 samples per timpoint; 27 samples total).  I did Western Blots on sets of 9 samples (1 of each genotype at each timepoint) for a total of 3 Western Blots.

I want to be able to compare samples within each timepoint to wild-type at that timepoint as well as potentially compare each sample to the 1-week wild-type sample.  I've done some reading and it seems like one-way ANOVA might be the way to go but I don't even know how to get started doing it.  For example, do I do the statistics on the raw densitometry numbers or the relative intensities (factoring in the control protein)? Any help is greatly appreciated.  I can post the actual numbers if needed.

Edited by NeuroGuy, 08 June 2011 - 12:55 PM.

[pcrman]

Yes, ANOVA is the right analysis to use. Because raw densitometry can be affected by many factors such as protein loading, exposure time, you may want to use expression levels relative to, for instance, wildtype, in the ANOVA analysis. You will have the mean and SD for each time point from the 3 samples.

[bob1]

Hmmm, ANOVA is for parametric samples where you can assume normal distribution of the data, you probably want a non-parametric equivalent such as the Kruskal-Wallace test.  You can do a post-hoc test such as Tukey's test on this data which will tell you which groups are significantly different.

Be aware that your results are going to be very very dependent on your ability to reproduce gels and antibody exposures exactly for your samples and control proteins.

As pcrman said, you will need to do the measurements relative to a loading control (in each lane).  The raw data will not work in my experience.