Hi everybody. I am currently using minfi for every analysis of Illumina 450k I am working on. But today I have run into what I consider a minor problem, i.e., more a matter of usability IMHO. Problem is, dmpFinder returns a data.frame with columns for the statistic, the p-value, the q-value and the intercept. And, if I would like to know which CpG's are hyper- or hypo-methylated from that table, I think I couldn't (I am talking about the two-category case here). For that, I would need the value of the second coefficient of the model, the actual slope of the fit. Besides, I would need to know the order of the levels in the factor used as a phenotype indicator (remember two categories case here), in order to interpret the sign of the slope. And minfi is executing: pheno <- factor(as.character(pheno)) which makes the previous ordering of the levels go away. I understand the general behavior of dmpFinder, and why its result is the way it is, but shouldn't it be a good idea to act differently in this case? I think that comparison of two groups, and splitting between hyper and hypo-methylated probes is very common. For now, I think I'll do the t-test, correct with p.adjust, and get the values I need on the fly. Regards, Gustavo --------------------------- Enviado con Sparrow (http://www.sparrowmailapp.com/?sig)

These are good comments. We are in the process of finishing up how we think we should look for DMRs, but I should fix this anyway. Might take a little while, but watch the devel version. Kasper On Wed, Nov 7, 2012 at 10:33 AM, Gustavo Fern?ndez Bay?n <gbayon at="" gmail.com=""> wrote: > Hi everybody. > > I am currently using minfi for every analysis of Illumina 450k I am working on. But today I have run into what I consider a minor problem, i.e., more a matter of usability IMHO. > > Problem is, dmpFinder returns a data.frame with columns for the statistic, the p-value, the q-value and the intercept. And, if I would like to know which CpG's are hyper- or hypo-methylated from that table, I think I couldn't (I am talking about the two-category case here). For that, I would need the value of the second coefficient of the model, the actual slope of the fit. > > Besides, I would need to know the order of the levels in the factor used as a phenotype indicator (remember two categories case here), in order to interpret the sign of the slope. And minfi is executing: > > pheno <- factor(as.character(pheno)) > > which makes the previous ordering of the levels go away. I understand the general behavior of dmpFinder, and why its result is the way it is, but shouldn't it be a good idea to act differently in this case? I think that comparison of two groups, and splitting between hyper and hypo-methylated probes is very common. > > For now, I think I'll do the t-test, correct with p.adjust, and get the values I need on the fly. > > Regards, > Gustavo > > > --------------------------- > Enviado con Sparrow (http://www.sparrowmailapp.com/?sig) > > _______________________________________________ > Bioconductor mailing list > Bioconductor at r-project.org > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor