Error with 2 component Deconvolution using TCGA-LUAD tumors/normals

[amartin2016]

I get the following error with running DeMixT_GS on 177 TCGA-LUAD tumors and 5 normal lung tissue samples. The R Code is attached.
Error: vector memory exhausted (limit reached?)

R Code: PhaseI_LUADtumor <- DeMixT_GS( data.Y=mrnaLUADTumor33SE, data.N1=mrna5LUADnormalSE, data.N2 = NULL, niter = 10, nbin = 50, if.filter = TRUE, filter.sd = 0.7, ngene.selected.for.pi = NA, nspikein = NULL, mean.diff.in.CM = 0.5, tol = 10^(-5), pi01 = NULL, pi02 = NULL, nthread = parallel::detectCores() - 1)

TCGA-LUAD Deconvolution 9_27_21.docx

[pengyang0411]

Dear user,

Thanks for your interest in our work.

I just went through your code, I don't see any problems. Can I ask how many genes did you input for this function? I suggest you run it on the server since the GS function is very computationally cost. It is very hard to run large-scale datasets on a personal PC.

Thanks, Peng

[amartin2016]

Hi Peng,

The samples each contained 60,000 genes. I’ve tried running it on a High Computing Cluster through Tufts University as well with the same problem and it takes longer to run the code typically. I’m successful when running GS function on TCGA-PAAD 178 samples, 4 normals all containing 60,0000 genes.

Thanks, Alex

---

Alexander S. Martin, MD Fellow, Hematology and Oncology Tufts Medical Center P 617.636.4580 @.**@.>

From: pengyang0411 @. Sent: Tuesday, September 28, 2021 12:40 AM To: wwylab/DeMixT @.> Cc: Martin, Alexander @.>; Author @.> Subject: [EXT] Re: [wwylab/DeMixT] Error with 2 component Deconvolution using TCGA-LUAD tumors/normals (#17)

EXTERNAL MESSAGE - TREAT LINKS/FILES WITH CARE

Dear user,

Thanks for your interest in our work.

I just went through your code, I don't see any problems. Can I ask how many genes did you input for this function? I suggest you run it on the server since the GS function is very computationally cost. It is very hard to run large-scale datasets on a personal PC.

Thanks, Peng

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/wwylab/DeMixT/issues/17#issuecomment-928791696, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AV2FEW33XHJ5WUWOQETD2O3UEFBKJANCNFSM5E4IRKBA. Triage notifications on the go with GitHub Mobile for iOShttps://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Androidhttps://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub. The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Tufts Medical Center HIPAA Hotline at (617) 636-4422. If the e-mail was sent to you in error but does not contain patient information, contact the sender and properly dispose of the e-mail.

Please consider the environment and the security of the information contained within or attached to this e-mail before printing or saving to an insecure location.

[pengyang0411]

Hi Alex,

I see, I think the problem is that there are too many genes and it requires large memory to compute the Hessian matrix. Therefore, I suggest you try DeMixT_DE function first, or select around 5,000 genes ranked by the standard deviation for normal reference from low to high, and then try DeMixT_GS function.

Please let me know if you have more questions, Peng

[amartin2016]

Thanks Peng,

Can you clarify how I use the output data from DeMixT_DE to inform how I proceed to running DeMixT_GS? My output from DeMixT_DE gives 1500 gene names and the pi gives 527 samples with piT and PiN1. I'm not sure what the importance of these 1500 genes are? Am I supposed to filter out the rest of the 59,000 genes in the next steps and only use the 1500 from DeMixT_DE?

Regards, Alex

---

From: pengyang0411 @.> Sent: Tuesday, September 28, 2021 3:29 PM To: wwylab/DeMixT @.> Cc: Martin, Alexander @.>; Author @.> Subject: [EXT] Re: [wwylab/DeMixT] Error with 2 component Deconvolution using TCGA-LUAD tumors/normals (#17)

EXTERNAL MESSAGE - TREAT LINKS/FILES WITH CARE

Hi Alex,

I see, I think the problem is that there are too many genes and it requires large memory to compute the Hessian matrix. Therefore, I suggest you try DeMixT_DE function first, or select around 5,000 genes ranked by the standard deviation for normal reference from low to high, and then try DeMixT_GS function.

Please let me know if you have more questions, Peng

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/wwylab/DeMixT/issues/17#issuecomment-929558952, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AV2FEW7AH74OVU6IRUZLHX3UEIJP5ANCNFSM5E4IRKBA. Triage notifications on the go with GitHub Mobile for iOShttps://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Androidhttps://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub. The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Tufts Medical Center HIPAA Hotline at (617) 636-4422. If the e-mail was sent to you in error but does not contain patient information, contact the sender and properly dispose of the e-mail.

Please consider the environment and the security of the information contained within or attached to this e-mail before printing or saving to an insecure location.

[pengyang0411]

Dear Alex,

The 1,500 genes from DeMixT_DE are selected based on differential expression with normal reference and mixed tumors. I suggest you set ngene.selected.for.pi = 5,000 in DeMixT_DE function, so the DeMixT_DE will give you 5,000 genes. Then you will filter the rest of the genes, and only use these 5,000 genes for the DeMixT_GS function.

Let me know if you have more questions, Peng