I will soon be receiving Illumina HiSeq 2500 data (150bp PE stranded reads). In the past we have been using Trinity as our assembler of choice, but it uses a default kmer value of 25. I believe this may have been optimal for GAIIx data (35bp) but is it still ok to use such a kmer for 150bp data?

All inputs will be greatly appreciated!


  • $\begingroup$ what sequencing are you doing- WGS/RNAseq/... ? $\endgroup$ – WYSIWYG Jun 3 '14 at 9:27
  • $\begingroup$ I am doing RNA-seq of multiplexed samples in order to determine differential expression between different samples $\endgroup$ – user2439719 Jun 3 '14 at 9:32
  • $\begingroup$ I have't used trinity.. I use cufflinks and for longer reads (>50nt) this is not really necessary (by default off in cufflinks for reads >50nt).. $\endgroup$ – WYSIWYG Jun 3 '14 at 9:36
  • $\begingroup$ As far as I know, Tophat-cufflinks require a reference, don't they? I am working on a non-model organism so will be doing de novo assembly, hence the use of Trinity. What do you mean is not necessary? $\endgroup$ – user2439719 Jun 3 '14 at 9:43
  • $\begingroup$ Oh okay.. In that case I cant say. I have some acquaintance with velvet but not trinity. In tophat (I am sorry I said cufflinks) that is used for enhanced prediction for splice junctions. For de-novo assembly the choice of k-mers should be such that it maximizes your N50 but does't consume a lot of resources. You can see this post $\endgroup$ – WYSIWYG Jun 3 '14 at 9:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.