There are different methodologies that you can use, but they depend of several factors such as:

1) Do you have a reference ? If you do, you can use (Multiple Sequence Alignment) MSA based correction methodologies. Unfortunately the only program that uses this methodology, Coral (http://www.cs.helsinki.fi/u/lmsalmel/coral/) doesn't work with fastq color space reads, but you can align the reads with BWA, and convert the SAM/BAM output into Fastq using Picard tools (http://picard.sourceforge.net/command-line-overview.shtml#SamToFastq).

2) If you don't have a reference you will need to use K-spectrum or suffix tree/array based tools such as Quake or Reptile. Nevertheless the only tool that works with color space format is Hybrid-SHREC (http://www.cs.helsinki.fi/u/lmsalmel/hybrid-shrec/).

For more comprehensive explanations about read correction error you can take a look to Yang et al. 2012 (http://bib.oxfordjournals.org/content/early/2012/04/06/bib.bbs015.full).

Good luck,

Hi Aureliano, that was really helpful (again :) ). Unfortunately there is no reference and what I believe is the best thing to do is to align the mate-pair reads to unitgs made with high quality reads (using BWA) and then extract the reads and correct with your proposed approach (MSA) and then use it in the scaffold.

Thanks

Luis

Hi Luis,

Before do that I'll try Hybrid-SHREC and I'll repeat the assembly. Some assemblies are improved after the error correction.

Good luck,

Best regards

High error rates with SOLiD, Really? They Always claim to be the most acurate NGS technology. What does the sequenceing centre have to say about this? I would have demanded to have the samples run again!

Hi Luis

I agree with Aureliano Bombarely and I hope his suggestions would work. I also have a solid 5500xl in my lab. You did not mention the size of genome and depth of sequence. In case of denovo assembly a deeper coverage will be required. Rarely people use Solid color space data for de-novo assembly due to multiple reasons for instance short reads and absence of mainstream assembling tools using colour space data.

Best of luck

Hi Luis,

I would get some HiSeq2500 paired-end coverage to build contigs and use SOLID mate-pair data only at scaffolding stage.

Dear Victor,

What software would you use for the scaffolding step with SOLiD mate pairs?

Thanks,

Anelda

Hi Luis,

For error correction you can try SAET tool, developed especially for SOLiD color space reads. This is free, but I cannot find the link now. You can send me your mail and I'll forward a distrib to you.

And I agree with others that the only option to get more or less satisfactory contigs from SOLiD is to make assembly in color space

Cheers,

Stepan

Hi Anelda,

I mapped SOLiD MP data against contigs with BWA0.5.10 and extracted info on pairs with unambiguous mapping to TSV file. Then used SSPACE for scaffolding. There might be other ways too :)

Victor

Hi Victor,

Thanks for your reply. I'm about to embark on a de novo assembly using Illumina & SOLiD data. We also have a 5500 and Torrent in-house and have until now not really known which tools to use for combined de novo assemblies. Do you have any gotchas/caveats you can warn me against or any trips or tricks that you can share to make this easier?

I'll appreciate any advice you or anyone else can give.

Best,

Anelda

We recently used a reference assisted assembly approach (using a genome from a related species in the scaffolding step) for a de-novo assembly using SOLiD data:

http://uu.diva-portal.org/smash/record.jsf?searchId=2&pid=diva2:563641&rvn=7

Cheers,

/Robert

Hi Robert,

Thanks for the link. In the thesis it is mentioned that you used base-space SOLiD data. How did you convert your SOLiD data to base-space?

Anelda

Anelda, the conversion to nucleotide space was done at the sequenceing facility. So unfortunatelly I am not completely sure about the details. This has rather recently become possible with the latest version of the SOLiD technology.

SOLiD reads are only good for scaffolding, we used them for our date palm genome assembly. But 3X can really do very little for the scaffolding.

Hi Robert and Rohan,

Thanks for the replies!

@Robert - I guess they used ECC then - that provides a FASTQ file with base space data. We have not had great success with the ECC module but I have to admit that the species sequenced had very high GC content, so it may be a contributer to the poor assembly we got using the base space data.

@Rohan - thanks for the info, I'll look into it. As far as I understand there is a very big reason not to convert to base-space. The problem begins when you have a single colour error. Because you use the last base + the current colour, everything downstream from your first colour error in your read might be converted incorrectly. I've been cautioned very strongly against converting colour data to base space because of this. Any comments/advice?

Thanks!

Anelda

Dear all, just to let you know a bit more about the developments with the SOLID data and the new problem that I am struggling with.

I finally could get a nice genome reconstruction using Illumina pe (150bp) data. Then I used the illumina genome reconstruction to map SOLID reads with PASS aligner (which in fact I am impressed how fast and accurate it seems to be, I also tried novoalignCS and BWA but PASS was just very convenient (80 % of mapping using raw cs data)). Using the pairing module in PASS I could retrieve the unique mapped pairs (with FF orientation). Everything worked beautifully until this stage. Now I have two questions about the strategy to follow.

1. Should I convert now to basespace and use the data in a convenient de novo pipeline, let 's say soapdenovo, is there any advantage in going back and use all the data?

2. or shall I just go forward and use the mappings for scaffolding?

I tried do the second one with SSPACE but unfortunatelly the sam/bam to SSPACE tab script uses suffixes and these paired sam files from PASS pairing module do not have suffixes, second read name comes after first read name and that's it.

I would appreciate any thoughts

Thanks

P.S, JUst to let you know ECC data maps far worst than colorspace. Although if I convert it to colorspace it maps better, something is really wrong with the ECC data.