Segmentation Fault

[Courtney Leisner]

Hello,

I am also experiencing an issue with a segmentation fault error. I am trying to run SMuFin on a potato genome that is ~ 850 Mb. The normal and tumor fastq files are both below 2 GB each.  I am running this on a server that has 24 processors and a total of 256 GB of RAM (one node). I have no issue running the test data on my server. I also had no problem indexing the reference genome with BWA. I am using BWA/0.7 or higher.

Below is my code:

mpirun --np 10 SMuFin --ref /data/run/cleisner/BRAG/SMUFIN/sim_523_data/ref_genome/523_asm_K71_PE_MSU_ILL_MP_ILL-scaffolds_newheader_filter_500bp --normal_fastq_1 /data/run/cleisner/BRAG/SMUFIN/sim_523_data/normal_fastqs_1.txt --normal_fastq_2 /data/run/cleisner/BRAG/SMUFIN/sim_523_data/normal_fastqs_2.txt --tumor_fastq_1 /data/run/cleisner/BRAG/SMUFIN/sim_523_data/tumor_fastqs_1.txt --tumor_fastq_2 /data/run/cleisner/BRAG/SMUFIN/sim_523_data/tumor_fastqs_2.txt --patient_id potato_523_genome --cpus_per_node 10

This is the error:

[cub:16915] [ 1] SMuFin(_Z24get_sequences_from_blockP10_TreeBlockPPcPm+0x1e) [0x469be8]

[cub:16915] [ 2] SMuFin(_Z15get_breakpointsPcRSt6vectorIP10_TreeBlockSaIS2_EES5_+0x35c) [0x469ff8]

[cub:16915] [ 3] SMuFin(_Z6masteriPPc+0xd1c) [0x46be06]

[cub:16915] [ 4] SMuFin(main+0xab) [0x46cef3]

[cub:16915] [ 5] /lib64/libc.so.6(__libc_start_main+0xfd) [0x3ab5e1ed5d]

[cub:16915] [ 6] SMuFin() [0x465af9]

[cub:16915] *** End of error message ***

--------------------------------------------------------------------------

mpirun noticed that process rank 0 with PID 16915 on node cub exited on signal 11 (Segmentation fault).

I have tried running this with more or less CPUs per node, but no luck. Any help would be great.

Thanks,

Courtney

[montse]

Hello,

Can you tell me what is the coverage of your samples?

Regards,
Montse

[Courtney Leisner]

Montse,
For the 'normal' samples I used a library with 7x coverage of my genome. For the tumor samples I made simulated data sets using wgsim from the reference genome.

Thanks,
Courtney

[montse]

Hi Courtney,

SMuFin has been designed to analyse samples in human, so we do not know what could be the behavior in other spicies. Even so, the 7x coverage is a very low coverage, and SMuFin default parameters have been adjusted for common sequencing scenarios (30x coverage or higher).

Regards,
Montse