RMA Nimblegen Data

[jmwhitha]

Excuse me. I meant to say in step 3
3. For cells which contain values less than 50, put 50. It's similar to something done by Chhabra et al. 2006 (doi:10.1128/JB.188.5.1817–1828.2006).

[jmwhitha]

Hey SequencingJelly,

Thanks for your reply. I was actually just about to implement the pipeline I arrived at today. Perhaps you can comment on it.

My planned steps are:
1. For each array, find the average and standard deviation of the random probes.
2. Subtract the average from each of the coding/non-coding probe values of the corresponding array.
3. For cells which contain negative values, put 50. It's similar to something done by Chhabra et al. 2006 (doi:10.1128/JB.188.5.1817–1828.2006).
4. Perform outlier detection using the SSDR method in He and Zhou 2008 (doi:10.1128/AEM.02536-07).
5. Replace outliers with the average of the remaining replicate probes.
6. Perform log2 transformation.
7. Perform quantile normalization.
8. Perform a median polish.

This workflow is an RMA-like normalization. Does that sound good? Any suggestions.

[SequencingJelly]

Hi Jason,

Did you get what you needed? I know a bit about RMA and NimbleGen data.

[jmwhitha]

Hello SeqAnswers community,

I'm rather new to analyzing microarray data, so please forgive my ignorance. I have a couple of questions.

Can I RMA normalize the probe data in my excel worksheet which contains on average 7 probes per gene and 3 technical replicates for each probe? If so, how? I have JMP Genomics. Not sure if I can do it with that.

Before finding out that RMA is one of the most common methods for converting probe level data into transcript abundance, I was going to Loess normalize the data then average all of the probes and technical replicates to get a single transcript abundance value for each transcript (corresponding to genes and intergenic regions). From what I've read though, it doesn't seem like people do this. Rather most people seem to RMA normalize their data and then average the technical replicates. Is that assessment correct?

I also have the Nimblegen array data files (e.g. .pair files) if it is necessary to go back to that level, and I found the following page which has a pipeline for this using R and bioconductor. I'm just not very good with these and I have limited access to a linux OS right now, so if there is an easier way with my Excel spreadsheet, I'd like to know that way. http://akka.genetics.wisc.edu/sandbo...c767cf222044f4

Thank you,
Jason