Dear pinky_pinkpinky, please see my comments below. Sincerely, Lukas > On 1 Mar 2017, at 17:54, pinky_pinkpinky [bioc] <noreply@bioconductor.org> wrote: > > Activity on a post you are following on support.bioconductor.org > User pinky_pinkpinky wrote Question: Error raised during execution of MEDIPS package: > > > Dear Users, > > I want to perform methylation analysis of 4 samples > > 1. normal and treated sample of methyl cystosine > > 2. normal and treated sample of hydroxy methyl cystosine. > > I want to perform both pair-wise comparison between 2 samples in each group and also across the groups. A comparison across 1. and 2. might be problematic as these are two different marks measured by two different antibodies. You can do a comparison of methylation in normal and treated samples (1.) Here, you have to put normal to MSet1 and treated to MSet2. It seems you do not have an Input/ background control (no enrichment) so do not use ISets. You can do the same thing for 2. separately. > > After executing MEDIPS.createSet on 4 samples i executed the below mentioned command > > CS = MEDIPS.couplingVector(pattern = "CG", refObj = T_mc) > > mr.edgeR_MC = MEDIPS.meth(MSet1 = T_mc, CSet = CS, ISet1 = N_mc, p.adj = "bonferroni", diff.method = "edgeR", MeDIP = T, CNV = F, minRowSum = 10) > > calculating genomic coordinates... > Creating Granges object for genome wide windows... > Preprocessing MEDIPS SET 1 in MSet1... > Calculating calibration curve... > Performing linear regression... > Intercept: -4.65438434638176 > Slope: 3.48363931263853 > Calculating relative methylation score... > Preprocessing INPUT SET 1 in ISet1... > No differential coverage will be calculated- only one group of MEDIPS SETs given. > Creating results table... > > mr.edgeR_HMC = MEDIPS.meth(MSet1 = T_hmc, CSet = CS, ISet1 = N_hmc, p.adj = "bonferroni", diff.method = "edgeR", MeDIP = T, CNV = F, minRowSum = 10) > > The above command raised the below mentioned error. and If I make MeDIP = false it works.? Why is it so. and is it correct to make it false in such studies? > > Calculating genomic coordinates... > Creating Granges object for genome wide windows... > Preprocessing MEDIPS SET 1 in MSet1... > Calculating calibration curve... > Error in MEDIPS.calibrationCurve(MSet = MSet1[[i]], CSet = CSet, input = F) : > The dependency of coverage signals on sequence pattern (e.g. CpG) densities is different than expected. No linear model can be build, please check the calibration plot by providing the MSet object at ISet. > > > Inorder to calculate differential methylation across 4 samples, I executed the below mentioned command > > > mr.edgeR_group = MEDIPS.meth(MSet1 = T_mc, MSet2 = T_hmc, CSet = CS,ISet1 = N_mc, ISet2 = N_hmc, p.adj = "bonferroni", diff.method = "edgeR", MeDIP = T, CNV = F, minRowSum = 10) > > Calculating genomic coordinates... > Creating Granges object for genome wide windows... > Preprocessing MEDIPS SET 1 in MSet1... > Calculating calibration curve... > Performing linear regression... > Intercept: -4.6543 > Slope: 3.48363 > Calculating relative methylation score... > Preprocessing MEDIPS SET 1 in MSet2... > Calculating calibration curve... > Error in MEDIPS.calibrationCurve(MSet = MSet2[[i]], CSet = CSet, input = F) : > The dependency of coverage signals on sequence pattern (e.g. CpG) densities is different than expected. No linear model can be build, please check the calibration plot by providing the MSet object at ISet. > > Again if I make MeDIP= false, It works. > > Question > > 1. Will it be correct to set the MeDIP argument to False , where I want to identify the differentially methylated region? When comparing different conditions you might not require CpG density normalisation and it is ok to set MeDIP=F. Please see also section 6.7 of the vignette. However, CpG density normalisation failed so there might be something wrong with the enrichment. I recommend to check the data in a genome browser and also examine the CpG calibration plot. > > 2. How to decide refObj value in MEDIPS.couplingVector when I have 4 methylated samples of 2 groups. If all samples are processed in the same way (same genome, same set of chromosomes, same window size), it does not matter which reference to chose. However, please see my notes above abt the experimental design. > > Inputs will be highly appreciated. > > Thanks!! > > > > > > Post tags: medips > > You may reply via email or visit Error raised during execution of MEDIPS package >