Subgroup GH (Growth hormone releasing) pituitary adenomas by microarray

[sotellme]

Subgrouping expression profiles of GH pituitary adenomas

Hello Everybody,

I am going to subgroup GH (Growth hormone releasing) pituitary adenomas according to their expression profiles by microarray and RT-qPCR. I do have 50 GH pituitary adenomas to subgroup and i wonder should i run all these 50 adenomas in the microarray to check their expression profiles or should i only take for example 5 of them to run in the microarray since they probably must have almost the same expression profiles? You know if i run all the 50 adenomas then it would take a lot of time. :yuck: What do you usually do when you have to subgroup a group of related things like in this case GH (Growth hormone releasing) pituitary adenomas?


Thanks for any input.

 

[Moonbear]

I have very little experience with microarray, but it seems to me that if the purpose is to find subgroups, it's those little differences that you're looking for. If you have 50 to work with, and don't know a priori which belong to what subgroups, I can't see any reason why running only 5 of them would make sense. What if by randomly selecting 5 of them, you ended up picking all 5 from the same subgroup?

Research does take a lot of time. I realize this is a big undertaking, and expensive too, but it's best to run them all together to reduce systematic errors than to run 5 now and the rest later. Make sure you have proper controls built in so you don't end up doing this twice! Do you have normal pituitaries to include? You'd want to be able to distinguish what's different in the adenoma from normal pituitary I would think. I assume you'll only do the RT-qPCR to confirm your findings in the microarrays, so you shouldn't need to do that for more than the most interesting genes and a few controls.

What's your hypothesis? You need to be clear about what your hypothesis is so that you can then include the proper controls.

 

[sotellme]

Thanks! I hope this is the way it is.

 

[sotellme]

Now i have statistical problems.

I have the problem in that i don't know which test i should use for my experiment. My experiment is that i have 50 tumours of GH releasing pituitary gland which i compare the expression profiles with 5 normal pituitary glands. After i have confirmed the fold increase or decrease of for example 3 interesting genes by RT-qPCR then what statistical test should i use to confirm these results? Note, i have not subgroup the different tumours yet all i have now is the expression profiles of 50 tumours of 3 interesting genes. What to do next to confirm these 3 genes expression profiles in these tumours by statistic?


Thanks very much for any suggestions!

 

[Monique]

You can do a t-test and associate p-values by using Excel. I believe the formula is (M1-M2) / (standard error * (M1-M2)) which is signal / noise.

You can also do the following:
Calculate a robust STDEV by throwing away 5% of the lower and higher distribution log ratio expression levels (so that you normalize for genes whose expression is not normal), with the remaining 90% you calculate STDEV: the 90% robust STDEV = p 0.5% Any gene that has a differential expression of > 2*STDEV or < -2*STDEV = significant.

I hope that makes sense.

 

[sotellme]

You can do a t-test and associate p-values by using Excel. I believe the formula is (M1-M2) / (standard error * (M1-M2)) which is signal / noise.

You can also do the following:
Calculate a robust STDEV by throwing away 5% of the lower and higher distribution log ratio expression levels (so that you normalize for genes whose expression is not normal), with the remaining 90% you calculate STDEV: the 90% robust STDEV = p 0.5% Any gene that has a differential expression of > 2*STDEV or < -2*STDEV = significant.

I hope that makes sense.

Thanks.

I have seen they normalize the Rt-qPCR data and then make a boxplot of it and also the median values in the boxplot. I wonder if they confirm the t-test based on the boxplot normalized values? This is the picture of it. Do you know why the values in the Y-axis are very low and how come i get them?

http://jcem.endojournals.org/content/vol86/issue7/images/large/eg0717616002.jpeg

p.S why can't we add images in this forum?