Hi Kathy Just quickly checked your document. Following is my answers to your questions: On Thu, Apr 12, 2012 at 9:06 AM, Hu, Xin <xin.hu at="" uth.tmc.edu=""> wrote: > Hi, Dr. Pan, > I am very impressive of Lumi package that you developed, can I have several questions regarding my application, please see the result attached, order of figure is according to your manual. > 1. I think my data has significant dye bias (fig2), right? After normalization(fig26,fig27),density of M-values is still imbalance(compared with your example on the manual), although CpG-site Intensity get consistent for all samples, how to solve this problem? The dye bias in your data is pretty typical. But not very severe. We have known that the M-value distributions of different samples can be very different. So normalization should not remove such difference. One advantage of the lumi package is that it does not directly normalize the data at the M-value level (since it violates the assumption of many normalization methods), instead we normalize the data at probe intensity level. Actually, your results show the normalization is pretty effective. For example, the within group difference becomes smaller while keeping the differences between groups.And the intensity distribution also become much more consistent. (We expect the intensity distribution should be very similar if there is no big copy number changes across samples). > 2. How can I export the adjusted data using Lumi to csv format, so that I could calculate beta_value eventually for publication, my boss asked me to use beta_value as most publication use beta_value. Is it possible to calculate beta_value instead of M-value by Lumi package? The MethyLumiM class basically is an extension of ExpressionSet class. So you can directly use "write.exprs" function to export the data, or you can extract the M-value using "exprs" function and then using "write.csv" function. To get Beta-value, you can run "estimateBeta" function > I appreciate so much for your great help! > Kathy > Pan