There are multiple methods supported by estimate.mean.fraglen. The SISSR algorithm is only the default. Did you try those? If you want to check the fragment length estimation, try plotting the coverage for the + and - strands separately, on top of each other. Check if the peaks line up after adjusting for the estimated insert size.
Michael On Fri, Oct 22, 2010 at 11:12 AM, [email protected] < [email protected]> wrote: > Dear bioc-sig-sequencing, > > If I run BioC2010 lab example, > http://www.bioconductor.org/help/course-materials/2010/BioC2010/Workflow.pdf, > I get the 184 to 196 bp average fragment length, as indicated on page 3 for > "ctcf" sample. > > If run same code on some of my own chipseq Arabidopsis samples > (Eland-aligned "export" files), for some, get average fragment length of > approximately 86 bp, for others get approximately 70 bp. Submitters of > these samples believe the average fragment length for these sonicated > samples is in the range of 200-250 bp. > > Looking at the ref paper for how the SISSR algorithm works (Jothi et al., > NAR 36 5221 2008), it looks like fragment length estimate not affected by > what DNA binding protein one may be studying, thus, perhaps fairly straight > forward calculation. > > So how to proceed? Perhaps just resize to the length found by the > estimate.mean.fraglen chipseq function, and proceed to analyze for which > peaks are different in two samples? > > Or is there an alternative option one should consider with the seemingly > too short average fragment length calculation results? > > > Thanks, > P. Terry > [email protected] > > [[alternative HTML version deleted]] > > _______________________________________________ > Bioc-sig-sequencing mailing list > [email protected] > https://stat.ethz.ch/mailman/listinfo/bioc-sig-sequencing > [[alternative HTML version deleted]] _______________________________________________ Bioc-sig-sequencing mailing list [email protected] https://stat.ethz.ch/mailman/listinfo/bioc-sig-sequencing
