Running abyss-pe

[VNou]

Hello, i am running ABySS with a fragment library of Staphylococcus
aureus that i found in this site: http://gage.cbcb.umd.edu/data/index.html.
I am running abyss-pe with the following parameters:
c=50 k=45 j=16 n=10 in='frag_1.fastq frag_2.fastq' , where frag_1 and
frag_2 contain the fragment libraries of the site. Every time i use a
different coverage (40,50,60,100). The site that i post contains also
jumping library. Should i incorporate and if yes how. Will the running
without the jumping libraries result in not good outputs, and with the
manner i run it is it good to have decent results? I am new in this
field so i want to ask to make sure that i am not doing all the things
wrong. Thanks a lot

[Shaun Jackman]

Hi,

Here are the recipes that the GAGE team used for the assembly:
http://gage.cbcb.umd.edu/recipes/abyss.html

You shouldn’t need to set the parameter c. ABySS will try to choose a good value for you. There’s no harm in experimenting though.

You can run the assembly without the jumping library, but you’ll get better scaffolding with the jumping library. To incorporate the jumping library, set
abyss-pe lib=frag mp=short frag='frag_12.cor.fastq’ short='short_12.cor.fastq'

Cheers,
Shaun

[VNou]

Hello,
i want to ask something more.
1)In the recipes that the gage team used the incorporation of the parameters is done with this line
ib='frag short' frag=frag_12.cor.fastq short=short_12.cor.fastq   and you wrote me lib=frag mp=short frag='frag_12.cor.fastq’ short='short_12.cor.fastq' . Is it the same thing?
 
2)I tried to assemble the human chromosome 14 by gage site with the use of 96 cpus and it was interrupted with this error:
The minimum coverage of single-end contigs is 5.91667.
The minimum coverage of merged contigs is 5.91667.
Building the suffix array...
Building the Burrows-Wheeler transform...
Building the character occurrence table...
sort: write failed: /tmp/sortwlczVi: No space left on device
error: `frag-3.hist': No such file or directory
make: *** [frag-3.dist] Error 1
make: *** Deleting file `frag-3.dist'

what is the problem? I used the recipe that is written in the site.

3) Also with 128 i had an error:
srun: First task exited 60s ago
srun: task 0: running
srun: tasks 1-127: exited
srun: Terminating job step 414895.0
slurmd[r01c3b5]: error: *** STEP 414895.0 KILLED AT 2012-05-28T20:14:49 WITH SIGNAL 9 ***

What is the problem? Ca you help me to find it out?

4)Is possible that sometimes the assembly procedure finish without a -stats file? And why some times the N50(or other results) in the output file is different from the stats file?
For example i had in the outfile:
Mateless         0
Unaligned   589819  33.8%
Singleton   557194  31.9%
FR             393  0.0225%
RF          284753  16.3%
FF            1271  0.0728%
Different   313605  18%
Total      1747035
n       n:200   n:N50   min     N80     N50     N20     max     sum
4476    53      3       201     173750  322274  947958  947958  2955212

and in the -stats file :

n       n:200   n:N50   min     N80     N50     N20     max     sum
5106    450     60      200     6087    14568   27937   53367   2824361 64t33k-unitigs.fa
4611    187     27      201     14363   32862   69045   108827  2947183 64t33k-contigs.fa
4476    52      3       201     173790  321962  947004  947004  2946210 64t33k-scaffolds.fa
~                                                                                          
why there is a difference

5)You said that abyss will try to choose a good value for c. Why at the end of the output file sometimes is written you should increase c parameter suggesting a input value?

Thanks a lot
Vaios

[Shaun Jackman]

Hi Vaios,

1. No. Those two commands are different. The command I suggested uses the short fragment library only for scaffolding and not for sequence assembly.

2. sort: write failed: /tmp/sortwlczVi: No space left on device
Your tmp disk is full. Use a different tmp disk, for example
export TMPDIR=/var/tmp

3. I’d guess that your scheduler killed the job for some reason. Perhaps it ran too long or was using too much memory.

4. In the output file, Ns are counted in the statistics, whereas in the final ${name}-stats file, only the nucleotides ACGT are counted.

5. You can ignore the message suggesting that you increase c. This suggestion is a rather old feature and is not really accurate anymore.

Cheers,
Shaun

[Shaun Jackman]

Hi Vaios,

I’ve not run into a segfault at this point in the program before. Try increasing the stack size by setting
ulimit -s unlimited
and run the command again.

Cheers,
Shaun

On 2012-05-29, at 7:27 , Vaios Noutsos wrote:

> Ok i fix the location of the temporary files exporting TMPDIR and then i execute the same example again. Although i have this output as error:
> Mateless 0
> Unaligned 22024 1.83%
> Singleton 163326 13.6%
> FR 28823 2.4%
> RF 0
> FF 28 0.00233%
> Different 988331 82.2%
> Total 1202532
> make: *** [128t31k-4.path2] Segmentation fault
>
> What can caused the segmentation fault? How can i fix it?
> Thanks
> Vaios
>
>
> 2012/5/28 Shaun Jackman <sjac...@bcgsc.ca>
> Hi Vaios,
>
> 1. Yes, your command line is correct. Illumina mate-pair reads produce chimeric reads, and so shouldn’t be used for sequence assembly.
>
> 2. Yes, you have to set TMPDIR in your job file. Try df -h /tmp /var/tmp to check if /var/tmp has more space free than /tmp, or are ask your systems department.
>
> Cheers,
> Shaun
>
> On 2012-05-28, at 14:55 , Vaios Noutsos wrote:
>
> > For
> > 1) If i have a short and a long jumping library with the command you suggested i have to write :
> > mp='short long' short=short_12.cor.fastq and long=long_12.cor.fastq? Using them in the scaffolding will save memory from the assembly process?
> >
> > 2) I have to make an export in the jobfile? Can i use the example tha you make(TMPDIR=/var/tmp)?
> >
> >
> > 2012/5/28 Shaun Jackman <sjac...@bcgsc.ca>