de novo assembly including Illumina and 454 paired-end reads

[ssully]

There's a bunch of hybrid de novo assembly threads but not sure I'm finding the answers to these questions, specifically about paired-end reads from two platforms (Illumina and 454).

454's 'paired end' are really mate pair ends from a 3-20kb fragment, swapped right to left in the read, both ends remain in 5--3' orientation (thanks to circularization), and are separated by a linker sequence.

5' right>-----[linker]left>------- '3

During assembly, Newbler splits these reads and also does quality trimming. It 'knows' that X bp should occur between the two half-reads.


Illumina paired ends are separate reads from each end of a ~350--600 bp fragment

read 1
5' --~100nt of left - 3'
read 2
5' --~100nt of reverse complement of right -- 3'

These reads may require some adapter and quality trimming as well.


So, can Newbler (I have v3.0) 'understand' Illumina paired-end reads (i.e. know that they represent of span of X bp)? Can it do trimming of adapters and low-quality bases?

Alternately, are there assemblers that 'understand' Illumina paired ends, but can also understand 454 PE reads? That is , tjhey know that the read must be split, linker discarded, and spacing of X bp between them? Can any also do quality trimming of SFF files?


I have raw fastq and sff reads, as well as trimmed versions of the same fq and sff reads. I also have (raw or trimmed) single-end sff and Fastq sets to use in the assembly (most of the data are single-end). Wanting to know what assemblers need what input, to fully exploit hybrid paired end information (e.g. for making better scaffolds)... with the least preprocessing on my part.

[danwiththeplan]

MIRA4:

http://mira-assembler.sourceforge.net/docs/DefinitiveGuideToMIRA.html#chap_denovo

Can handle hybrid assemblies, and also can handle paired-end reads in the orientation you describe (there's an example of this in the link above)..

It's a bit of a memory hog though. You may require a high-memory system.

[ssully]

That helps a lot (for MIRA) , thanks!

So as long as I convert the 454 Paired End sff to Fastq with sff_extract, MIRA will process and assemble them correctly *as paired ends*? It still understands that those reads are 'paired end' and not single end? I'm curious as to how it does that -- does it create an interleaved Fastq or two separate -1 and -2 fastq files?

[danwiththeplan]

It still understands that those reads are 'paired end' and not single end?

My understanding is yes

I'm curious as to how it does that -- does it create an interleaved Fastq or two separate -1 and -2 fastq files?

Actually don't know.

MIRA4 works by using a manifest file that defines the data to go into the program.

look at section 5.3.3. Manifest for data sets with paired reads (in the link above).

There's a parameter called segment_placement that defines how the pairs are arranged (ie >> or <> or >< or << or whatever) and (I think) the expected separation.

As for separate FASTQ files for left and right reads, I think MIRA expects Illumina data to be this way, but I don't know how 454 data works.In the example in the link above the data is defined as 454 data and only one file is given, so maybe you don't have to split the pairs. not sure about this one.

[ssully]

The example in the mulitple platform manifest *seems* to indicate the 454 PEs can be left as one fastq file. The insert size and SD need to be provided (which I can do) or 'autorefine' to let MIRA figure it out. Using 'autopairing' would mean not even having to tell MIRA the direction of the reads in a pair.


Looks good, but I'll write the author and see if I can get a clearer view.

[wanghao]

I'll write the author and see if I can get a clearer view.

Can you also post the reply here if you get it from the author.

[flxlex]

So, can Newbler (I have v3.0) 'understand' Illumina paired-end reads (i.e. know that they represent of span of X bp)? Can it do trimming of adapters and low-quality bases?

Yes, Newbler will figure out the pairing from the fastq files, provided the read IDs conform to the 'standards' (see the fastq entry on wikipedia). No, you cannot tell Newbler the span, as it figures this out for itself, regardless of where the data came from. It maps pairs and determines the mode and stdev of the distribution based on that.

Newbler will trim low-quality bases. It can do adaptor trimming through the -vt flag (you have to provide it with a fasta file of adapter sequences, probably both in forward and reverse complement orientation).

[ssully]

Can you also post the reply here if you get it from the author.

That was the wrong way to do it -- instead I've joined the MIRA user forum, which is what the author recommends.

[bastianwur]

With that data, I'd use GapFiller with both the reads, and throw the filled up 454 contigs together with the Illumina data into an assembler which can take both, like e.g. Ray.