[ians]

i'm curious, how are you exporting 454 LIMPs to be compatible with SOAP.
We had trouble converting to "Illumina style" LIMPS.

[guyleonard]

i'm curious, how are you exporting 454 LIMPs to be compatible with SOAP.
We had trouble converting to "Illumina style" LIMPS.

SOAPdenovo seems to quite happily take fasta and fasta qual files, so I just used sff_extract to create those...

python sff_extract -l titanium_linker.fasta -s output.fasta -q output.qual -x output.xml input.sff

and then in the SOAPdenovo config file I have something like this;

# 3kb library
avg_ins=3000
reverse_seq=0
asm_flags=3
rank=2
f=input.fasta
q=input.fasta.qual

BTW, what exactly do you mean by LIMP?

Oh it dawns on me that you probably mean a mate-pair library...

That is more tricky - you could go through the sff_extract but dump it to a fastq file instead and then use a sed command to convert the .f .r to the more useable /1 /2 of Illumina and dump the .fn to a separate file,

However, this -454-pairs- set of scripts, AFAIR, should do it automatically...took me ages to find this but it's very useful! Just checked, you still have to convert .f anf .r in each file to /1 or /2 but at least forward and reverse reads are separate this way rather than combined in one file...

[ians]

BTW, what exactly do you mean by LIMP?

Yes, exactly, Long Insert Mate Pair. Some people refer to this library type as "Jumping Pairs" as well. LIMP libraries are made through fragment circularization. Otherwise, end sequencing on a fragment is called a Short Insert Paired End (SIPE).

# 3kb library
avg_ins=3000
reverse_seq=0
asm_flags=3
rank=2
f=input.fasta
q=input.fasta.qual

This is where your soap performance may be getting killed. When pairs are in one (Fasta) file, use 'p='. Also, SOAP doesnt actually use quality information; the 'q=' takes fastq files that have reads and quality information. Hence, there may also be something funky going on here since you can't specify two different sequencing libraries for one soap-cfg library. What is the log output during scaffolding of this library? I imagine there will be very little pairing for this library.

[guyleonard]

This is where your soap performance may be getting killed. When pairs are in one (Fasta) file, use 'p='.

Ah yes, ahem (the manual isn't the most helpful).

This increased N50 from 4729 to 21188 with just a quick run. That seems much more like it. Time for some serious calibration!

[ians]

Yeah the documentation is a bit rough. This manual is a bit better.

You're doing a fungus right? How much sequence coverage did you get for each LIMP library?

Depending on the sequencing coverage of the LIMP libraries, you might have to tweak the scaffolding parameters as well. You may have to lower pair_num_cutoff and pair_num_cutoff for the 20k library to get more connections during scaffolding.

[Wallysb01]

Neat, I hadn't seen that program I had tried a little bit with SSPACE and TGNET but never really got anywhere that dramatically reduced %N.

SOAP has a stand-alone-ish gap closer (http://soap.genomics.org.cn/about.html#resource2) that you could try. I've seen some people get pretty good results from iteratively repeating a stand alone gap closure and scaffolding post de novo assembly.

Ah, I see now someone brought this up already, sorry for the duplication.

[flxlex]

Late reply, but anyway: looks like the pseudo-reads in newbler with the 454 LIMPS (or whatever you call them) works best. But, man do you get a lot of 'N's... Is the DNA from a heterozygous individual? In that case, have you tried the '-het' option, which potentially could close some gaps? And the SOAP gap closer, or other such program, may be worth trying.

Have you tried Celera? We have good success using it on 454 data, now trying to add Illumina (raw) reads. Adding raw illumina reads to newbler apparently tends to fragment contigs (I have seen that too), I recall them (454) saying they are working on that.