Newbler de novo assembly and repeats

[wiart]

Hi,

I'm using Newbler de novo assembler from the command line (runAssembly), with a set of 18,000 reads.
I'd like to tweak the "repeat" threshold since Newbler estimate that >98% of the reads are repeats, and I end up with 0 contigs...

Here is the command:
runAssembly -ss 12 -sl 16 -sc 1 -ml 25 -mi 80 -o ./newbler seqs.fa

BTW, do somebody have information about how this "repeat detection" step work? I couldn't find much about that in the GS manual... (the manual mention a -rst repeat score threshold, but only for assembly to a reference, with runMapping command)

Thanks,
Laurent

[jnfass]

Just a few thoughts (no direct answers) ...

I'm sorry - I can't help you with tweaking the assembly parameters (such as -rst) as I've never used anything but the defaults ...
But, have you looked at your sequences? Maybe newbler's right ...

I often look for unusually frequent sequences in a quick and dirty way -- something like:
cat seqs.fa | grep -v ">" | cut -c1-25 | sort | uniq -c | sort -rn -k1,1
But that'll only tell you if there you have contamination at the beginnings (or in the same positions, if you cut other characters) in every read (of course you miss the ones on the other strand).
More specifically to repeats - have you tried blasting to a repeat library?

[alex_dl]

I found the same kind of behavior in reads with vector sequence.
Be sure that you provide a trimming sequence if the library was constructed with a vector (as in a fosmid library)