Assembling 454 and illumina reads from an eukaryotic virus

[Monir]

Really excited to post the first question in this forum. I have been browsing the topics for several days and finding a lot of useful information. My thanks to everyone making this such a resourceful place!

I am currently working with some illumina and 454 data generated from a viral genome that infects a eukaryote. The illumina reads I have are paired end reads. I am looking for an efficient way to assemble this virus genome. I ran all the illumina data through CLC genomic workbench for a test assembly. However, it generated a lot of contigs (about 3,000,00). Some of them are pretty big ranging from 84kb to 2kb. But most of them are small contigs of several hundred base pairs. A priliminary BLAST search revealed that a lot of these contigs probably originated from mitochondria/chloroplast, which means our sample had contamination from the host. I have checked several posts and looks like there is not an efficient way of assembling the data from different NGS platforms together. I have also found that students/scientists use Velvet and other free assembler softwares, which are pretty good.

My specific questions are:

1. What kind of workflow you recommend to assemble the data I have? Should I go for assembling the 454 and illumina data together, or assemble them separately? (Please provide some detailed information).

2. Which software (other than CLC or Seqman) do you recommend for the work? Velvet, Newbler....and so on..

3. How can I get rid of the reads that are actually contamination from the host? Other than BLASTing in NCBI, is there any particular tool to facilitate their elimination?

4. I believe the paired end information of illumina data will help me scaffolding once I finish the assembly. Do you think I am right? Is there any software that can use my assembled contigs and paired end information to scaffold them?

Sorry for such a long post, but I really need some help to clear up my confusion. Your inputs will definitely save me a lot of time and will help to avoid the pitfalls. I guess some of the members here have experience with working on NGS data of viruses. If they can share their advice, I will be really grateful.

And forgive my ‘not-so-scientific’ description of the problem…I am fairly new in the field! Ha ha.

[swbarnes2]

3) Alignment. You need a software like bwa or bowtie. Make a reference genome of the host genome, plus your virus, and align all your reads to that. Then, make a filtered .bam that contains the lines that map to virus and all unmapped reads, then assemble that with velvet or whatever. Some software, like velvet will even take the .bam as input, otherwise, you'll need to convert that filtered .bam back into a fastq first. There are a few programs that can do this, Picard is one of them.

[Monir]

Hi,

Thanks a lot for the info. Really appreciated!!

I don't think we have any particular reference sequence for the virus I am assembling. This is a 'unique' virus, let's say! Ha ha.

In this case, if I use the host genome (and the chloroplast + mitochondria genome too) as a reference sequence and just exclude the reads that map to it for further assembly, will that be a good idea? A lot of times viruses tend to pick up sequences from their host. If I follow this approach, I might lose some sequences which are actually present in the host. What do you think? I guess, I will need to discuss this issue with my Professor.....

[maubp]

Plant virus sequencing eh?

2. Which software (other than CLC or Seqman) do you recommend for the work? Velvet, Newbler....and so on..

MIRA http://www.chevreux.org/projects_mira.html is good for smaller genomes and does hybrid assemblies of mixed sequence technologies.

3. How can I get rid of the reads that are actually contamination from the host? Other than BLASTing in NCBI, is there any particular tool to facilitate their elimination?

As mentioned above, you could map your reads to the host genome (or at least in this case the host chloroplast and mitochondria - try those from the closest available plant if yours is not sequenced) and remove those reads which map. This assumes you don't expect fragments of any similar virus to already exist integrated into the host genome.

[Wallysb01]

Do you have predominately one type of sequence or another? You could first assemble just one type of data then use programs designed more for genome refinement or reference guided reassembly with the other type or both types sequentially.

I.e. Pagit: http://www.sanger.ac.uk/resources/software/pagit/

There are others that I don't remember, a bit of time on google/pubmed would find them.

Oh and I almost forgot, you could put your host's (and anything else you might be picking up) sequence as a database for seqclean to screen your reads (pre-assembly) or post-assembled contigs. http://sourceforge.net/projects/seqclean/