Newbler2.3 & cDNA assembly

[moinul]

Hi all,

I am running newbler (v2.3) cDNA one-step assembly for about a week using 04 shotgun files, ~2.4 GB each. It seems just hanging over at detangling alignment phase for over 03 days. The status are shown as below-
---------------------------------------------------------------------------------------
Setting up long overlap detection...
-> 3142131 of 3142131, 3130211 reads to align
Building a tree for 15374046 seeds...
Computing long overlap alignments...
-> 3130211 of 3130211
Setting up overlap detection...
-> 3142131 of 3142131, 1032064 reads to align
Building a tree for 29430104 seeds...
Computing alignments...
-> 3130211 of 3130211
Checkpointing...
Detangling alignments...
-> Level 2, Phase 8, Round 1...
---------------------------------------------------------------------------------------
I used 08 cpus and at present it takes 01 cpu and 49% memory. Can anyone tell me would it complete at all or what should I do now?

Moinul

[sklages]

2.3 is "old". You should upgrade to Newbler 2.5 as it is known to work a lot better for cDNA assemblies. In summer 2.6 will follow.

hth, Sven

[moinul]

thanks hth, Sven, I also tried with Newbler 2.5 cDNA assembly but couldn't resolve the problem arose there. I think if you see my older thread "Newbler de novo assembly" posted on 28-05-2011, you can help me regarding this ..........

Moinul

[sklages]

Oh, well OK, I see.
I got this error as well reported it to Roche and it is still not resolved.
I wanted to assemble a non-normalised cDNA library with some transcripts having a really high coverage. I just "reduced" the dataset using 'cd-hit-est', some kind of "in-silico-normalisation" ;-) Reducing the coverage of certain transcripts helped, though this approach is not applicable to all kind of experiments (depending on the questions you want to answer).

Sven

[flxlex]

Have you removed ribosomal RNA sequences present in the data (e.g. by adding them via the -vs option)? That might also help.

[moinul]

thanks flxlex, but how do I get the rRNA sequences? simply download the fasta databases and mention with -vs flag?

And why should I remove them? they are also a part of my transcriptome ie. total RNA. Would it reflect the actual transcriptome size if I exclude them?

Moinul

[flxlex]

thanks flxlex, but how do I get the rRNA sequences? simply download the fasta databases and mention with -vs flag?

I would try to get some rRNA sequences specific for your organism, or from closely related ones. Don't use the entire database, that would slow things down considerably...

And why should I remove them? they are also a part of my transcriptome ie. total RNA. Would it reflect the actual transcriptome size if I exclude them?

They are represented to such a high coverage, that they cause problems for the whole assembly. Are you using total RNA? Then they make up the majority of your sample, I guess... Doing an assembly with a fraction of the reads will get you the rRNA sequences of your organism (these can also be used for the filtering, BTW), and the lengths of these you can add to the remainder to get total transcriptome size.