Velvet paired reads not working properly

[Admiralenola]

Dear all,

I'm having a problem doing de novo assembly of paired reads in Velvet. (I am working on a bacterial, single chromosome genome.) The contigs assembled are extremely short, having an n50 of just 25, max 272! If I only use one of the fastq files (in other words, not using paired reads), I get contigs with an n50 of around 20.000.

Naturally, I'm assuming that either I'm doing some of the input wrong, or my reads are not really what I think they are.

I have 100nt reads from a HiSeq 2000 machine. The reads come in two different fastq files. I have already used the reads for SNP analysis by aligning them against a reference genome, and that works nicely. (I did that in MAQ.)

I've remembered to flag my reads as shortPaired when running velveth, and I've used shuffleSequences_fastq.pl to shuffle the reads. I've tried with a variety of insert sizes. (Alignments to reference suggests the insret is around 200) I've tried to reverse complement zero, one and both of the read files. (We've had radio silence from our sequencing centre, so unfortunately I don't really know how they prepared the data)

My sequence data seems properly sorted; Here are the first reads from each file:

@HWI-ST790:2:1101:1481:2166#TAGCTT/1
ATGAGGCTGGAATGCCTTCAAAGCGACTTCGCTGCGGATTCTCCAGCGCCTTGGCCTGGTTCCACTTCGTGATGATTTCATCGATGGATGCCGGAAAATG
+HWI-ST790:2:1101:1481:2166#TAGCTT/1
aaaece`cgggggiiiiihiiiiiiiihiiiiiihiifhhaggihiiigggebeeddccdcccdccccccccccd_acbcdbcc`]_ccbcccccXacbb

@HWI-ST790:2:1101:1481:2166#TAGCTT/2
ATCAGGCCAGTTTGTCATCGATGATGTGGTGCGCGGCCTCACCGAAAAAATGACGCGGCGTAATGCGCATGTTTTCAAGCCGGATGGGAGTTTGCAGGAG
+HWI-ST790:2:1101:1481:2166#TAGCTT/2
_b_eeeeegggggiiiiiiihhiiihiihiiiiiihiiiihihihiigfeddddcccccccccccccccccccccccccccccaacccbc^bbccdcccc

My questions are:
1. Is this a known problem?
2. Is there a tool that takes in a set of your reads and tells you what they are? (Orientation, insert size, etc)

Thanks in advance!

[ians]

I suspect you are feeding velvet the reads properly; Velvet expects that pairs be in the same file and in forward-reverse order. It's a good idea to read the velvet manual:

4.1 Input sequence les
Velvet works mainly with fasta and fastq formats.
For paired-end reads, the assumption is that each read is next to its mate
read. In other words, if the reads are indexed from 0, then reads 0 and 1 are
paired, 2 and 3, 4 and 5, etc.
If for some reason you have forward and reverse reads in two dierent FASTA
les but in corresponding order, the bundled Perl script shueSequences fasta.pl
will merge the two les into one as appropriate. To use it, type:
> ./shuffleSequences_fasta.pl forward_reads.fa reverse_reads.fa output.fa
There is also a shueSequences fastq.pl script for Sanger-style FastQ les.
Concerning read orientation, Velvet expects paired-end reads to come from
opposite strands facing each other, as in the traditional Sanger format. If you
have paired-end reads produced from circularisation (i.e. from the same strand),
it will be necessary to replace the rst read in each pair by its reverse complement
before running velveth.

[Admiralenola]

Sorry if I did not make this clear in the original post, but I have done exactly that. So, after using the shuffleSequences_fastq.pl script, reads are lined up next to their "mate". For example, the first two reads in this resultant, shuffled file are:

@HWI-ST790:2:1101:1481:2166#TAGCTT/1
@HWI-ST790:2:1101:1481:2166#TAGCTT/2