Hi Shankar,

Another suggestion: instead of bedGraph, try the more concise fixedStep wiggle 
format, and wigToBigWig, as described here:

https://lists.soe.ucsc.edu/pipermail/genome/2010-December/024318.html

Hope that helps,
Angie

----- "Galt Barber" <[email protected]> wrote:

> From: "Galt Barber" <[email protected]>
> To: "Hiram Clawson" <[email protected]>
> Cc: [email protected], "Shankar Ajay Subramanian" <[email protected]>
> Sent: Tuesday, March 15, 2011 7:54:42 PM GMT -08:00 US/Canada Pacific
> Subject: Re: [Genome] bedGraphToBigWig memory issue
>
> Yes. I was thinking that perhaps
> a call or two to needlargemem
> should have called needhugemem
> instead.
> 
> However since we could not reproduce it thar makes it hard to be
> sure.
> 
> -galt
> 
> On Mar 15, 2011, at 6:16 PM, Hiram Clawson <[email protected]>
> wrote:
> 
> >
> > Go into the source tree and edit the file: src/lib/memalloc.c
> > change the definition of the maxAlloc variable:
> >
> > static size_t maxAlloc = ...
> >
> > Currently it allows 16 Gb in a single allocation.  Add a factor
> > of 4 to make it 64 Gb.
> >
> > --Hiram
> >
> > ----- Original Message -----
> > From: "Shankar Ajay Subramanian" <[email protected]>
> > To: "Galt Barber" <[email protected]>
> > Cc: [email protected]
> > Sent: Tuesday, March 15, 2011 5:32:41 PM
> > Subject: Re: [Genome] bedGraphToBigWig memory issue
> >
> > Hi Galt,
> >
> > Thanks for your response. I updated the executable yesterday with
> what
> > appeared to be a new version (updated Mar 14 on the UCSC website)  
> > but I
> > still end up getting a similar error:
> >
> > needLargeMem: trying to allocate 34345569624 bytes (limit:  
> > 17179869184)
> >
> > I forgot to mention that the program writes a partial bigWig file  
> > (roughly
> > 6GB) before aborting.
> >
> > Let me know if you need additional information for debugging/ 
> > diagnosis.
> >
> > Shankar
> >
> > On Tue, Mar 15, 2011 at 8:23 PM, Galt Barber <[email protected]>  
> > wrote:
> >
> >>
> >> Hi, Shankar
> >>
> >> I tried to reproduce this error, but was unable to.
> >> I used a 34GB bedgraph file and had no trouble.
> >>
> >> Since this problem may have been fixed somewhat recently,
> >> I suggest you try downloading a fresh version
> >> of bedGraphToBigWig and run it again.
> >>
> >> -Galt
> >>
> >>
> >> 3/14/2011 8:29 PM, Shankar Ajay Subramanian:
> >>
> >> Hello,
> >>>
> >>> I'm trying to create a per-base coverage track from whole-genome
> >>> sequencing
> >>> data and I opted to use the bigWig format to display this on the 
> 
> >>> UCSC
> >>> browser. I downloaded the 64-bit linux 'bedGraphToBigWig' binary 
> 
> >>> utility
> >>> to
> >>> do so (Feb-18 version).
> >>>
> >>> The original read data are in BAM format (200GB), which I then  
> >>> converted
> >>> to
> >>> bedGraph (49GB). When I tried to convert intermediate bedGraph  
> >>> file to
> >>> bigWig I end up getting the following error even though the  
> >>> process runs
> >>> on
> >>> a machine with>100G RAM:
> >>>
> >>> needLargeMem: trying to allocate 34345569624 bytes (limit:
> 4294967296 
> >>> )
> >>>
> >>> From talking to our sys admin it appears that the utility calls a 
> 
> >>> library
> >>>>
> >>> where a hard-coded memory limit is specified for a 64-bit machine
> >>> (memalloc.c). Is there any way to get around this? I considered
> >>> down-sampling but didn't know if that would help since the  
> >>> bedGraph file
> >>> is
> >>> likely to contain nearly as many data-points with Column4 values 
> 
> >>> being
> >>> lower.
> >>>
> >>> Appreciate your help.
> >>>
> >>> Best,
> >>> Shankar
> >>>
> >>> Postdoctoral fellow
> >>> NIH/NHGRI
> >>>
> >>
> >>
> > _______________________________________________
> > Genome maillist  -  [email protected]
> > https://lists.soe.ucsc.edu/mailman/listinfo/genome
> _______________________________________________
> Genome maillist  -  [email protected]
> https://lists.soe.ucsc.edu/mailman/listinfo/genome
_______________________________________________
Genome maillist  -  [email protected]
https://lists.soe.ucsc.edu/mailman/listinfo/genome

Reply via email to