velvet assembly with non-unique reads: collapse or not?

mgg · 02-24-2012, 11:56 AM

Does the level of sequence duplication in Illumina reads affect the way velvet works?

I'm trying velvet to assembly some bacterial Illumina data. It's not doing so well, producing a lot of small contigs rather than fewer larger ones.

The data are 10-105 nuc reads that have been processed to remove adapters using cutadapt (hence the size range)

The nominal coverage (read length x number of reads)/genome length is reasonable (range: 21-243, median 100).

However the data all have quite a lot of sequence duplication (according to FastQC (range: 23-80; median 58)

Using Velvet Optimiser identifies a kmer between 69-71,but the output is still lots of smaller contigs.

Would reducing that duplication level help velvet? If so what software would anyone recommend (not FastX; its writer told me it wasn't designed to today's read lengths - which is why it couldn't clip adapters in my libraries)

thanks, look fwd to everyone's suggestions, and have a great weekend

m

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
How to collapse sequences excluding sequencing artifacts? (454)	Yevaud	Bioinformatics	3	01-30-2013 07:59 AM
[Velvet,assembly] core dumped occured by runnning velvet	matador3000	De novo discovery	0	12-17-2011 08:31 AM
Velvet assembly	katussa10	De novo discovery	2	12-12-2010 11:22 AM
Regarding Unique reads, Unique alignments	sridharacharya	RNA Sequencing	2	09-20-2010 06:39 AM
Velvet de novo assembly of Solid reads HOWTO	KevinLam	De novo discovery	1	01-10-2010 01:11 AM

02-24-2012, 11:56 AM	#1
mgg Member Location: London, UK Join Date: Nov 2011 Posts: 12	velvet assembly with non-unique reads: collapse or not? Does the level of sequence duplication in Illumina reads affect the way velvet works? I'm trying velvet to assembly some bacterial Illumina data. It's not doing so well, producing a lot of small contigs rather than fewer larger ones. The data are 10-105 nuc reads that have been processed to remove adapters using cutadapt (hence the size range) The nominal coverage (read length x number of reads)/genome length is reasonable (range: 21-243, median 100). However the data all have quite a lot of sequence duplication (according to FastQC (range: 23-80; median 58) Using Velvet Optimiser identifies a kmer between 69-71,but the output is still lots of smaller contigs. Would reducing that duplication level help velvet? If so what software would anyone recommend (not FastX; its writer told me it wasn't designed to today's read lengths - which is why it couldn't clip adapters in my libraries) thanks, look fwd to everyone's suggestions, and have a great weekend m