velveth: Can't malloc

[ninn...@gmail.com]

Dear all,

I am running JAFFA in the assembly mode and run into some problems:


velveth: Can't malloc 103 chars: Cannot allocate memory
Running Assembly for k=19
Running Assembly for k=23
velvetg: Could not write to directory_23/Log: No such file or directory
oases: Could not write to directory_23/Log: No such file or directory
Running Assembly for k=27
velvetg: Could not write to directory_27/Log: No such file or directory
oases: Could not write to directory_27/Log: No such file or directory
Running Assembly for k=31
velvetg: Could not write to directory_31/Log: No such file or directory
oases: Could not write to directory_31/Log: No such file or directory
Running Assembly for k=35
velvetg: Could not write to directory_35/Log: No such file or directory
oases: Could not write to directory_35/Log: No such file or directory
velveth: Could not open directory*/transcripts.fa: No such file or directory
Oases K-mer=19 assembly failed
Oases K-mer=23 assembly failed


Any ideas, what s going on here?

I call like this:

JAFFA/JAFFA-version-1.06 run JAFFA_assembly.groovy fastq_folder

Thanks!

[Nadia Davidson]

Hi,

Sorry for the slow reply. It sounds like the velvet assembler has run out of memory, but it's unusual for it to happen while running velveth. I've only encountered a problem when it runs velvetg. Do you know if your system memory is maxing out while you run JAFFA?

I'm not sure if this will help, but you could try to run JAFFA with just a single k-mer (e.g. 27) and see if that works. The command would be:

JAFFA/JAFFA-version-1.06 run -p Ks="27,28,1" JAFFA_assembly.groovy fastq_folder

If it does work, you could then try running with progressively more k-mer (e.g. Ks="27,32,4" which corresponds to Kmers of 27 and 31. The assembly will be done with K-mers from the first to (second number -1) in increments of the third number.

You should still have okay power to detect fusions with just a single k-mer, but the more the better. Alternatively, if you have longer reads, you could just run the Direct mode and avoid assembling altogether.

Cheers,

Nadia.

[rohen...@gmail.com]

Hi Nadia,

I'm resurrecting this thread, because I'm running into the same problem with velvetg. It just doesn't run, because it's trying to allocate too much memory.
I see that you ran into this problem as well, so I was wondering if you found a way to solve it?

Thank you.

Kind regards,
Romina

[Nadia Davidson]

Hi Romina,

Unfortunately assembly can use a lot of memory. Do you know how much velvet is requesting when it runs out and how much your system has? 

Some options are:

- Don't assembly (using JAFFA's direct mode if your reads are say 75bp or longer). If you have shorter reads (50bp) try a different fusion finder like SOAPfuse, STAR-fusion or pizzly.

- Try to use just a single k-mer for the velvet assembly (see my previous reply in this thread)

- Try to digitally normalise the data before running JAFFA  (e.g. using something like diginorm)

- Try using a server with more RAM.

Cheers,

Nadia.