A Problem with Short Reads in Newbler

[AMacFadyen]

Hi all,

Just looking for some help with performing de novo assemblies with Newbler.

We've sequenced our genome using Illumina and have a 200 bp paired end library. I want to perform a de novo assembly with the data using Newbler however it fails to assemble.

I'm uploading both .fastq files, Newbler will index them, acknowledge that I have selected pair-end and then "completes" the assembly but doesn't generate any contigs. The read status error says the reads are too short.

Given the read lengths are 30 bp they should be sufficient, though I think the preferred is 50. I've tried changing the parameters to compensate for 30 bp read lengths but with no luck.

Any feedback would be greatly appreciated.

Alison

[nickloman]

Why do you want to use Newbler? As the in-house assembler from Roche this is not a natural choice for Illumina data, except for when you are combining Illumina reads with 454 data. Even if it does work, it's likely suboptimal. I'd try a short read assembler instead, any of the ones listed here are probably reasonable choices, except ALLPATHS-LG which isn't compatible with your dataset: http://gage.cbcb.umd.edu/assemblers/index.html

[nickloman]

Also, prepare to be disappointed with the results of a de novo assembly generated with 30bp reads, even if paired.

[AMacFadyen]

I've got 454 data that is being used as a reference for 3 other genomes that were sequenced using Illumina. We have Illumina data for the reference genome as well and I've tried doing the assembly by combining this with the 454 and it still gives me the same error. Doing the assembly with the Illumina and 454 data for the reference would be handy but currently my priority is getting an assembly for the Illumina data sets (or at least one of them). I'd like to use just one assembler for all the data which is why I wanted to use Newbler.

I'll have a look at one of the ones listed as well. It'll give me something at least.

Thanks for the information.

[nickloman]

According to Lex Nederbragt's excellent blog, Newbler has a minimum length for reads of 50bp and a default minimum overlap of 40bp, which is probably the reason you don't have any luck combining the data. For hybrid Illumina/454 assemblies it might be worth reducing those values so the short reads are used (see http://contig.wordpress.com/2011/04/...-read-contigs/).

If you just want to use the Illumina reads, a de Bruijn graph based short-read assembler will likely perform much better (the overlap-layout-consensus method used by Newbler isn't tailored for large numbers of very short reads).