Newbler Warning - Primer contamination

[loba17]

Hello,
I am using Newbler 2.6 to do a de novo cDNA assembly with 454 reads. The program is giving me a warning about possible primer contamination (i.e. TGTTTTTTTTTCT). I checked the assembly and found about 250 contigs (out of 20,000 in total) with the reported primer sequence. We used the MINT cDNA synthesis kit and the reported sequence seems to be part of the MINT kit primers. I could use the -vt flag with runAssembly and provide a fasta file with the primer sequences to trim the reads but would this be correct?

Reason for not trimming:
> the primer is part of the mRNA and therefore should not me removed - I will loose some information

Reasons for trimming:
> the primer sequence could lead to incorrect asemblies
> the primer sequence might be part of the mRNA but not the protein - this region could cause false positives in blast searches

I know that RNAseq data have this characteristic bias (e.g. random hexamer primer) but I think nobody is trimming the read because of it. I could assemble the reads without trimming and remove the contigs with the primer sequence not at the end.

Is anybody willing to share his thoughts or experience on this? I would appreciate your help. Thanks!

[maubp]

I had something similiar, and got better assemblies removing the MINT primer.

[loba17]

Dear maubp,

thanks for the answer. I was wondering if you trimmed the reads or the contigs? Did you use the -v option or a different program for the read trimming? Would you mind specify "better assembly". How did you assign the quality improvement?

[sklages]

I had something similiar, and got better assemblies removing the MINT primer.

same here, we had a lot of EST libs created with MINT system; we always removed primer sequences (as I would do for any other libs as well).

[maubp]

I think I tried both the -v option (for Newbler) and trimming the reads (for Newbler and MIRA).

Without trimming the reads I got some very strange coverage patterns where at one end of a contig there was a MINT adapter that was overly represented. For a MIRA EST example see Figure 5 in Milne et al. 2012 http://dx.doi.org/10.1093/bib/bbs012

[loba17]

Dear sklages, dear maubp,

thanks for your help.

I was reading more about the trimming step. It seems that the -vt flag is the best way to proceed. People also recommend to use the -vs flag to remove rRNA sequences. For this part I could downloaded the RNAmmer fast file from the CBS website and used it with the -vs flag. This would, however, only cover prokaryotes. Any suggestion about eukaryotes rRNA sequences? I could get the ribosomal sequence from NCBI but I guess the file would be rather large.

In addition, I also found references recommending the -urt flag. But this seems to be controversial. I tried it and it resulted in a large number of contigs ... three times more than before. Therefore I think it is best not to use it at least for my assembly. Are there other reasons (not) to use it?

[flxlex]

Regarding the '-urt' flag: it is supposed to give more complete transcripts. It could be that there are ore contigs per isogroup ('gene'), but that you would have to check. So, maybe the flag is actually usefull (I would compare data with and without it - not just looking at contig number - to make sure)