MPI version of BIMBAM
Fan, Yiping
Hi, Yongtao,
How are you? I am trying to use BIMBAM to do the imputation for a group of patients conditioning on the references from HAPMAP.
Since we have many patients, I’ll try to use the parallel version of BIMBAM.
We have a Linux cluster in house. My system admin already installed the MPI version of BIMBAM. I have several questions for you.
Does MPI version of BIMBAM put each patient on a separate CPU or each reference to a separate CPU?
If MPI version of BIMBAM put each patient on a separate CPU, does that mean BIMBAM does not use the info from cohort/study population, i.e. treats each patient independently?
Does the MPI version require a different input file format?
Thank you very much for your help!
Yiping
Email Disclaimer: www.stjude.org/emaildisclaimer
Matthew Stephens
From: Matthew Stephens <steph...@gmail.com>
Date: Fri, Feb 5, 2010 at 12:16 PM
Subject: Re: MPI version of BIMBAM
To: yipin...@stjude.org
> Does MPI version of BIMBAM put each patient on a separate CPU or each
> reference to a separate CPU?
>
it puts each patient on separate CPU.
>
>
> If MPI version of BIMBAM put each patient on a separate CPU, does that mean
> BIMBAM does not use the info from cohort/study population, i.e. treats each
> patient independently?
>
during imputation it treats each independently (as does IMPUTE v1).
very little accuracy is lost by doing this.
But during association analysis of course all patients are considered together.
Matthew
>
>
> Does the MPI version require a different input file format?
>
>
>
> Thank you very much for your help!
>
>
>
>
>
>
>
> Yiping
>
>
>
> ________________________________
> Email Disclaimer: www.stjude.org/emaildisclaimer
>
> --
> You received this message because you are subscribed to the Google Groups
> "BIMBAM HELP" group.
> To post to this group, send email to bimba...@googlegroups.com.
> To unsubscribe from this group, send email to
> bimbam_help...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/bimbam_help?hl=en.
>
Yongtao Guan
> Hi, Yongtao,
>
>
>
> How are you? I am trying to use BIMBAM to do the imputation for a group of
> patients conditioning on the references from HAPMAP.
>
>
>
> Since we have many patients, I’ll try to use the parallel version of BIMBAM.
>
> We have a Linux cluster in house. My system admin already installed the MPI
> version of BIMBAM. I have several questions for you.
How many patients do you have?
>
>
> Does MPI version of BIMBAM put each patient on a separate CPU or each
> reference to a separate CPU?
Yes, it actually put panel individuals on different CPU (or processors
to be accurate) as well.
However, data are pooled across different CPU of course.
>
>
> If MPI version of BIMBAM put each patient on a separate CPU, does that mean
> BIMBAM does not use the info from cohort/study population, i.e. treats each
> patient independently?
We treat cohort individuals (patient in your case) independently by
design, not limited by the use of MPI.
>
>
>
> Does the MPI version require a different input file format?
It should be the same as current version, but read the earlier manual
to make sure.
http://quartus.uchicago.edu/~yguan/bimbam/resource/instruction.pdf
>
>
>
> Thank you very much for your help!
>
>
>
>
>
>
>
> Yiping
>
>
>
> ________________________________
> Email Disclaimer: www.stjude.org/emaildisclaimer
>
> --
> You received this message because you are subscribed to the Google Groups
> "BIMBAM HELP" group.
> To post to this group, send email to bimba...@googlegroups.com.
> To unsubscribe from this group, send email to
> bimbam_help...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/bimbam_help?hl=en.
>
--
Yongtao Guan, PhD
http://home.uchicago.edu/~ytguan/
Fan, Yiping
Thank you very much for the quick reply!
I have ~1000 white patients. Since it actually puts panel individuals on different processors, I guess the number of patients in each panel depends on the number of processors in our cluster
(-np option), correct?
Yiping
-----Original Message-----
From: Yongtao Guan [mailto:ytg...@gmail.com]
Sent: Friday, February 05, 2010 12:35 PM
To: Fan, Yiping
Cc: bimba...@googlegroups.com
Subject: Re: MPI version of BIMBAM
Fan, Yiping
Based on my previous experience, the imputed SNPs for each individual may be different if you impute them separately. We use the intersection of the imputed SNPs for association study. How do you pool the imputed SNPs in the MPI version of BIMBAM? Thanks!
Yiping
-----Original Message-----
From: Yongtao Guan [mailto:ytg...@gmail.com]
Sent: Friday, February 05, 2010 12:35 PM
To: Fan, Yiping
Cc: bimba...@googlegroups.com
Subject: Re: MPI version of BIMBAM