Velvet denovo

[bair]

Dear all,

We have an alignment file in bam format including three type of reads with different length ( pair-end and mate-pair), we want to do denovo assembly for the novel sequence by using velvet.

We got three fastq files with unmapped reads from the bam file by using picard ViewSam.jar.

How to input these reads to velvet? Since the reads in fastq file are not in proper order, e.g no pair-end relation, can we treat these reads as singletons?

Can we use velvet to handle the three fastq files together?

Many thanks!

[francesco.vezzi]

You are free to treat the reads as they are singletons but this will probably lead to a bad assembly with a lot of short contigs. If you have a way to keep the paired read information (with the SAM file it must be possible) this will help a lot the assembly phase.

As far as the last question on the on line manual you will find an example to how give to velvet more then one file in input

Francesco

[bair]

Thank you Francesco,

So pair-end information will lead to better assembly, I need to extract this from the sam file.

[maubp]

Sorting the SAM/BAM file by name may help - or you could do the sorting after converting to FASTQ. See also the thread "BAM to fastq how?" on the samtools-help mailing list this August:

http://sourceforge.net/mailarchive/f...=samtools-help

[Torst]

Originally Posted by bair

We have an alignment file in bam format including three type of reads with different length ( pair-end and mate-pair), we want to do denovo assembly for the novel sequence by using velvet.We got three fastq files with unmapped reads from the bam file by using picard ViewSam.jar.

So you only want to de novo the unmapped reads to find novel segments?

How to input these reads to velvet? Since the reads in fastq file are not in proper order, e.g no pair-end relation, can we treat these reads as singletons?

You should write a script to split your unmapped reads into pairs & singletons, and do this for each of the three libraries.
eg. lib1p.fa lib1s.fa lib2p.fa lib2s.fa lib3p.fa lib3s.fa

Can we use velvet to handle the three fastq files together?

Yes, but you may need to re-compile velvet to support CATEGORIES=4. Here is a suitable partial command line:

% velveth -shortPaired lib1p.fa -shortPaired2 lib2p.fa -shortPaired3 lib3p.fa -short4 lib1s.fa lib2s.fa lib3s.fa
% velvetg -exp_cov auto -cov_cutoff auto