Combining multiple observations of 1 gene for ANOVA in JMP Genomics
May 20, 2009 3:53 PM(1150 views)
I am working on analyzing an affymetrix expression CEL data set using JMP genomics. I have successfully inputted all the data and created a design file (which I hope is correct), normalized the data and examined it for outliers.
My experiment is as follows: 3 treatment conditions, 3 arrays/group for 9 arrays.
The issue is that when I do the oneway anova analysis the software is comparing each observation of the gene individually, rather than combining them.
To clarify, there are 27663 genes examined per array. However, there are numerous probes for each to maintain quality. I believe these should all be combined within each array before the between group comparisons are made. There should be 27663*9=248967 observations of genes, however because of the repeats there are ~830,000 entries.
I hope you have gotten the answer to this question, but if not, I hope this helps. I'm the product manager of JMP Genomics, and these forums are usually for user-to-user communication but I see no one has answered your question so I don't want to leave you hanging. If you ever need individual attention on questions, please don't hesitate to submit a request to JMP technical support. There is a journal under our documentation and help section about contacting tech support, and you can open a track via web at: http://support.sas.com/ctx/supportform/index.jsp
The One-Way ANOVA process in JMP Genomics performs only row-by-row analysis, so if you want to perform transcript-level analysis using it, you need to either: 1. summarize the data beforehand during import with the CEL import process, by selecting a method on the Summary tab. 2. use our Statistics for Rows process to create a summary data table post-import
To use Statistics for Rows, select the data table you were using as input to One-Way ANOVA. Leave the "Variables to be Summarized" box blank. Choose ProbesetID as your "Variable by which to Summarize." Select a method for summary (e.g., mean, median) on the Options tab. Then, use the output data set (with suffix _drs) as your input to One-Way ANOVA.
An alternative is to use our ANOVA process, which allows you to select a By Variable to summarize probes into a group for analysis. In your case you would select ProbesetID to do the analysis by probeset ID and summarize across all probes in the probe set.
Hope this helps. If not, please email me at Shannon.Conners@jmp.com.