de novo assembly of PROTON transcriptome data

[rasel@barj]

Hi all,

We performed a plant whole transcriptome sequencing in Ion PROTON. We have about 10 Gbp data of 74 million reads in a single run using v2 proton kits. Average read length is 129 bp. Which assembler will give us better output for de nove assembly. Could anyone suggest me?

Thanks
Rasel

[GenoMax]

Have a look at this thread: http://seqanswers.com/forums/showthread.php?t=1610

There are several other threads available (look at the top panel in "similar threads").

Short answer may be that you may have to try a few assemblers out. If you have a similar species for which a transcriptome/genome is available then you can use that as a rough reference to do some comparisons to the assembly that you build.

[rasel@barj]

Thanks GenoMax for your suggestion.

We tried newbler but it was failed to assemble the data. Now MIRA is running for 13 days but still not completed. We don;t know why it is taking too much time although we only took 5 Gbp data for MIRA assembly running now. Last time in case of 2 Gbp PROTON transcriptome data MIRA took 12 days to complete and the metrics were not good.

Actually I am worried about the data type of PROTON which may cause problem in assembler. Unfortunately there is no sufficient past example or I couldn't find specific post related to PROTON transcriptome (plant) data de novo assembly.


Thanks again!
Rasel

[colindaven]

Soapdenovo or in your case the -trans version has been pretty capable or at least quick in my hands.
CLC assembly cell (commercial) or CLC genomics workbench might also be worth a look.

Velvet might be an option too.

[rasel@barj]

Thanks colindaven for your suggestion

[jonathanjacobs]

CLC's assembler is quite good (and very fast) - so I would second that suggestion. Also, you may want to try IDBA's transcriptome assembler: http://i.cs.hku.hk/~alse/hkubrg/proj...ran/index.html

That being said -- 129 bp reads are not going to give you great assemblies - no matter what assembler you choose. Proton RNA-seq data is going to be much better for read mapping and mutation/SNP mapping. For de novo assembly of transcriptomes, you would do much better using PE reads.

[super0925]

Thanks colindaven for your suggestion

Hi rasel@barj ,
I have similar issue , average reads length is 114 bp at Proton , 28 M reads per run.
Proton use MIRA as their GUI plugin, but it seems that it is very slow from your post.
So which assembler did you use eventually ?!

[super0925]

CLC's assembler is quite good (and very fast) - so I would second that suggestion. Also, you may want to try IDBA's transcriptome assembler: http://i.cs.hku.hk/~alse/hkubrg/proj...ran/index.html

That being said -- 129 bp reads are not going to give you great assemblies - no matter what assembler you choose. Proton RNA-seq data is going to be much better for read mapping and mutation/SNP mapping. For de novo assembly of transcriptomes, you would do much better using PE reads.

Hi jonathanjacobs,
(1)Why do you say that? (i.e. 129 bp reads are not going to give you great assemblies)
(2)Why do you say you would prefer use PE reads ?
(3) Is the IDBA's transcriptome assembler suitable for Proton data as well? how does it compare with Abyss, Velvet or Soapdenovo?

[super0925]

How about Trinity?

[rasel@barj]

Hi super0925

We tried several assemblers but no one can build high quality assembly. Among all the assemblers Trinity gives much better metrics than others.