Increasing contig lengths

[arundurvasula]

I'm working on a project to identify by sequence viruses infecting grapevines.

I have single end Illumina reads (50 bp) and have been trying to assemble them using a combination of Velvet and PRICE. I've been able to get to a max contig length of around 1500 with Velvet and an n50 of 46. After putting this output through PRICE, I can increase the n50 to 195. However, I am having trouble increasing my contig length after this. Do you have any advice regarding contig extension with single end reads?

[Brian Bushnell]

Have you tried varying the kmer length when assembling? Also, it would be helpful to know more about your data, like the read length and total amount, and quality metrics.

I encourage you to read this thread:

http://seqanswers.com/forums/showthread.php?t=42555

[luc]

... and the amount of contaminating sequences?

Have you tried varying the kmer length when assembling? Also, it would be helpful to know more about your data, like the read length and total amount, and quality metrics.

I encourage you to read this thread:
http://seqanswers.com/forums/showthread.php?t=42555

[arundurvasula]

Thanks for the replies.

I used VelvetOptimiser to determine optimal k-mer length. Our data contains a mixture of grape and virus reads, but we removed the reads that aligned to the grape reference genome. Our read length is 50 bp and we have 7,764,190 reads after filtering out the grape reads.

Here is the quast output from the optimal velvet run:

All statistics are based on contigs of size >= 100 bp, unless otherwise noted (e.g., "# contigs (>= 0 bp)" and "Total length (>= 0 bp)" include all contigs).

Assembly contigs
# contigs (>= 0 bp) 3547
# contigs (>= 1000 bp) 1
Total length (>= 0 bp) 326445
Total length (>= 1000 bp) 1073
# contigs 941
Largest contig 1073
Total length 156559
GC (%) 46.84
N50 162
N75 122
L50 305
L75 584
# N's per 100 kbp 0.00

[mastal]

Are you trying to assemble genomic data or transcriptomic data?

What is the expected genome size of the virus genome you are trying to assemble?

What kmer length have you used?

As Brian already mentioned above, I would play around with the kmer length
when using velvet, to see what kmer length gives you the best n50.

Have you done any QC, adapter trimming or quality trimming on your reads?

[SNPsaurus]

Do you think the viral genome will be divergent within a sample from replication errors? That could cause issues for assembly if there are lots of related kmers at a location instead of just one or two alleles and a low level of sequencing error.

[arundurvasula]

I was able to assemble my data using IDBA_UD. I set it to cycle through k mers that were less than my read size and it produced a 15000bp sequence: idba_ud -r ../data/trimmed-reads/LV89-02.fa -o ../results/contigs/008 --mink 19 --maxk 49 --step 2

[jpummil]

quast quality stats of the assembly?

[arundurvasula]

All statistics are based on contigs of size >= 100 bp, unless otherwise noted (e.g., "# contigs (>= 0 bp)" and "Total length (>= 0 bp)" include all contigs).

Assembly contig
# contigs (>= 0 bp) 295
# contigs (>= 1000 bp) 37
Total length (>= 0 bp) 199556
Total length (>= 1000 bp) 86293
# contigs 295
Largest contig 15124
Total length 199556
GC (%) 46.27
N50 759
N75 426
L50 53
L75 141
# N's per 100 kbp 0.00