454assembly using Newbler: job time expectation?

[Jordy224]

Greetings

I would like to know how long should the following assembly to run using newbler?

The following are details:

gsAssebler
software version: 2.3
Data: 454 data
Input: 3 .sff files with a total of 1,434,170 reads
Genome: eukaryotic unknown size
Not pair end.

I am using the command line and not the GUI. The manual suggests that runAssembly is faster and more efficient. In my previous jobs, I find this to be true. ie.

shell> runAssembly -g , -cpu Num 4 ./input #command used.

Hardware specs:
CPU: 4 processors 2.66ghz ea
RAM: 16gb


The 454newblerprogress.txt file shows that it is currently “detangling alignemnts”. This is great but the file has not updated for more than 24hrs!

The job is running for 6 days, a deadline is looming...
Should I assume it the newbler stalled, died or just gave up?


Thanks.

[flxlex]

Long detangling points to a problem with the dataset. You could try the -large option which I think will solve it. It shortcuts some steps during detangling (and alignment).

If coverage is low, ie less than say 15x, also increasing alignment stringency could help: -mi 96 or 98, -ml 60, 80 or even 100.

Good luck!

[Jordy224]

great...
The assembly finished in 13 hours:
start: Sat Mar 27 19:21:59
stop: Sun Mar 28 08:25:07

What is the advantage of increasing the ml length or mi length with low coverage reads?
I am interested in doing additional reading on the topic and Im sure others have had the similar issues if coverage is low.

Also, I found that the N50 decreased by using large option and switching from -ml60/-mi90 parameters to default parameter (N50 = 775, N50=540).

I assume that low quality reads are dropped by setting the overlapMinMatchIdentity to a higher value, which increases contig quality. But I am unclear on this topic.

[flxlex]

Increased mi and ml prevents the assembler 'looking for overlap' where there is not enough coverage to actually have overlap. I don't think it drops low quality reads any more than default parameters, though...