Using DESeq with Biological Replicates with Different RNAseq Read Lengths

jbvt · 05-20-2013, 02:18 PM

Is it okay to use DESeq with count data from biological replicates having different RNAseq read lengths?

thanks

turnersd · 05-20-2013, 03:04 PM

I might consider making this an element in your design table, as was done with the pasilla library type variable in the DESeq vignette. But I'm not sure how necessary. Perhaps others can share their thoughts.

chadn737 · 05-20-2013, 05:24 PM

Does each replicate have different read lengths? If so, then I would certainly make this a factor.

Wolfgang Huber · 05-21-2013, 08:34 AM

To avoid biases by caused different efficiencies of the aligner in mapping reads with different length, one might shorten the longer reads to make them all have the same length.

Otherwise, I agree with Stephen and Chad(?). The ends will justify the means, and the usefulness of your result will depend on the relative sizes of the "batch effect" caused by your experimental design and the biological effect of interest.

Best wishes
Wolfgang

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
many biological replicates - 'traditional' statistics vs Cuffdiff or DESeq/edgeR?	sjm	RNA Sequencing	0	03-31-2012 09:42 AM
DESeq without biological replicates (dataset: Marioni et al.)	Andrea Apolloni	Bioinformatics	3	01-13-2012 05:58 AM
overdispersion and biological replicates	shilez	Bioinformatics	3	08-29-2011 08:43 AM
overdispersion and biological replicates	shilez	RNA Sequencing	0	08-25-2011 07:10 PM
Inconsistency between biological replicates	Nicholas_	Bioinformatics	1	04-06-2011 04:18 AM

05-20-2013, 02:18 PM	#1
jbvt Junior Member Location: blacksburg, va Join Date: May 2013 Posts: 1	Using DESeq with Biological Replicates with Different RNAseq Read Lengths Is it okay to use DESeq with count data from biological replicates having different RNAseq read lengths? thanks

05-21-2013, 08:34 AM	#4
Wolfgang Huber Senior Member Location: Heidelberg, Germany Join Date: Aug 2009 Posts: 109	To avoid biases by caused different efficiencies of the aligner in mapping reads with different length, one might shorten the longer reads to make them all have the same length. Otherwise, I agree with Stephen and Chad(?). The ends will justify the means, and the usefulness of your result will depend on the relative sizes of the "batch effect" caused by your experimental design and the biological effect of interest. Best wishes Wolfgang __________________ Wolfgang Huber EMBL

05-20-2013, 03:04 PM	#2
turnersd Senior Member Location: Charlottesville, VA Join Date: May 2011 Posts: 112	I might consider making this an element in your design table, as was done with the pasilla library type variable in the DESeq vignette. But I'm not sure how necessary. Perhaps others can share their thoughts.

05-20-2013, 05:24 PM	#3
chadn737 Senior Member Location: US Join Date: Jan 2009 Posts: 392	Does each replicate have different read lengths? If so, then I would certainly make this a factor.