Question: BitSeq: Required input for 3' bias correction on transcripts.
4.2 years ago by
Peter Glaus • 70
Peter Glaus • 70 wrote:
Hi Dave, regarding BitSeq's non-uniform read distribution model. BitSeq learns two separate bias models for sense and anti-sense reads so these reads are being distinguished. Kind regards, Peter. On 29/01/13 09:19, Dave Gerrard wrote: > Hi Peter, > Thanks for your reply. A lot of 'conventional' cDNA protocols created > double-stranded DNA, from which the orginal strand information was > lost but some RNA based library protocols retain the strandedness (see > http://biology.stackexchange.com/questions/1958/difference-between- strand-specific-and-not-strand-specific-rna-seq-data > ). The difference would matter most when you have overlapping > transcritps (common in small genomes) or you have no clue about > transcription strand (ncRNAs with no poly-A). > > However, given that I am working on poly-A extracted RNAs from the > sparse human genome, and BitSeq's mapping strategy, it's not too bad > an assumption that the coding strand was the source. > > The reason I asked was because Cufflinks refuses to do non-uniform > distribution without strand information (according to the manual). > From you answer, it sound's like BitSeq infers the strand from the > mapping direction (which is what I want) but does it then do something > different for reads that map in the anti-sense direction? > > Dave > > > On 27 January 2013 20:37, Peter Glaus <email@example.com> <mailto:firstname.lastname@example.org>> wrote: > > Hi Dave, > as far as I understand the RNA-seq protocols, the "endedness" of > reads is determined depending on whether they were generated in > first-strand synthesis or second-strand synthesis. > > Either way, this can be also determined based on transcripts > "strandedness" (in our case known from annotation) and read being > aligned to sense or anti-sense strand of the transcript. So BitSeq > tries to determine end for each read and it is safe to use the > non-uniform model. > > If you want to make sure that everything runs OK you can use > verbose flag to get more information and check the number of reads > that are reported in the pre-processing part and watch for any > unexpected warnings. > > Regards, > Peter. > > > > On 01/25/2013 12:02 PM, Dave Gerrard wrote: >> I have RNAseq data generated from poly-A capture RNAs. There >> appears to be a strong 3' bias to the mapped reads but the >> library prep was NOT strand specifc. I would like to correct for >> the positional bias in the mapped reads and Bitseq has a >> 'uniform' parameter which can be set to 'FALSE' if reads are not >> uniform. Please can you tell me if this requires 'stranded' >> mapped read information (as per the cufflinks manual) or whether >> It will work without knowing the source strand? >> >> Thanks, >> Dave Gerrard >> -- >> email@example.com >> <mailto:firstname.lastname@example.org> >> >> Bioinformatics | The University of Manchester | Michael Smith >> Building | Room B.1079 | Oxford Road | Manchester | M13 9PT >> T: 0161 275 5737 >> >> http://personalpages.manchester.ac.uk/staff/David.Gerrard/ > > > > > -- > email@example.com <mailto:firstname.lastname@example.org> > > Bioinformatics | The University of Manchester | Michael Smith Building > | Room B.1079 | Oxford Road | Manchester | M13 9PT > T: 0161 275 5737 > > http://personalpages.manchester.ac.uk/staff/David.Gerrard/ [[alternative HTML version deleted]]
ADD COMMENT • link •