Some proteins (abnormal proteins!?) expression pattern is highly different among individual sample so that it is very hard to get the significance difference between the group. For example, the band intensity (band area) of control group and Drug intervention for protein A expression in my study is as follows
Con, 1 : 3000
Con, 2: 2000
Con, 3: 500
Drug A, 1: 300
Drug A, 2: 200
Drug A, 3: 50
In this case although the protein expression in Drug-treated group is almost 10 times less than that in control group, there won't be significance difference between these two groups because of the variation of protein expression pattern despite using same amount of loading protein (loading control expresses in similar manner). In such case, how could we get the significance difference between the groups? Do you guys have any way to sort out such issues?
From the data you presented, it looks already like there is a significant difference between the controls and the drug-treated samples. You could improve on the statistics by collecting more data for both the control and the drug treatment. This will likely reduce the error bars on the average calculated from both sets of data.
Hi Bhakta,
While observing the values it looks like Drug works but because of the variance between replicates causes your t-test to be not significant. In the example you showed, the p-value would be p=0.0867. What you can do is log-transform your data either using natural log or log2 or log10
Using natural log transformation (LN) from your example it is now:
Con, 1 : 8.00
Con, 2: 7.6
Con, 3: 6.21
Drug A, 1: 5.7
Drug A, 2: 5.3
Drug A, 3: 3.91
When you do the t-test you would then report as log(p-value) = 0.039
Hi Bhakta,
there is not enough information in your question regarding your experiments. the main thing I would like to know is the linear range of your western blot - i.e. is 50 really 4 fold less than 200. for that you should have done a calibration curve for western first. are you measuring intensity or area? for example, HRP based western blots have 1-1.5 orders of magnitude in the linear range. did you check for lane overloading? band broadening will result in skewing of the quantification.
in short - I think your best bet is to properly assess the suitability of the method in question. (and if we're at it, why don't you run SRM/MRM experiment?)
Cheers,
D.
Hello Albert,
Thank you for your suggestion. Could you please elaborate it a little more how can I perform the statistical evaluation using LN methods? In fact, I have never used this methods for statistical analysis. Therefore, I will appreciate if you share the methodology to do this.
- Badges
- Science topic
- Similar topics
- Chemistry
- Pharmacology
- Pharmaceuticals
- Drugs