question about canopy.post()

[Liu, Delong (NIH/NHLBI) [E]]

Hi Yuchao,

 

I run Canopy for a few test runs on my data.  Here is what I have observed from the runs:

 

Run 1: K = 4:6, numchain = 2 or 5, canopy.post produced results without any error.

 

Run 2: K = 4:6, numchain = 20, canopy.post generated an error message “Error in config.summary[i, 1] : subscript out of bounds”.

For both tests, I used burnin=100 and thin=10.

Do you have any suggestions on what may cause the error message?  

I would be happy to send you the data I used, so that you replicate the results.

Thanks,

Delong Liu

10 Center Dr

Bethesda, MD 20892

301-451-2773 (w)

 

[Yuchao Jiang]

Hi Delong,

Thanks for your interest. To speed up computation in MCMC and to avoid autocorrelation between neighboring trees that are samples consecutively, Canopy saves samples trees with a skip. The error that you see from the output you sent me seems to be due to the fact that with the writeskip, burnin, and thinning you don't have any posterior trees left. I would suggest running your trees longer to get more accurate posterior evaluations. Another option is you can change the burn in and thinning parameter. Either way I would suggest looking at the likelihood plot generated for sampling.

Please also update your package from the GitHub page since we recently pushed some changes.

Please feel free to let me know if you have any further questions.

Cheers,

Yuchao

Sent from my iPhone

--
You received this message because you are subscribed to the Google Groups "canopy_phylogeny" group.
To unsubscribe from this group and stop receiving emails from it, send an email to canopy_phyloge...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/canopy_phylogeny/DFDED68A2B2F724AA157A2836706EB49237761D2%40msgb10.nih.gov.
For more options, visit https://groups.google.com/d/optout.

[Liu, Delong (NIH/NHLBI) [E]]

Yuchao,

 

Thanks for your quick response.  I will tune the burnin and thinning parameters with sampchain from canopy.sample.

 

The current version of Canopy I am using is Canopy_1.1.1, which is same as the one listed from the GitHub web site.

 

Delong

[ame...@b612.email]

Hi Yuchao,

You suggest running your trees longer to get more accurate posterior evaluations, the longer trees mean bigger K or bigger numchain?

Sincerely,

Amelie

在 2016年11月2日星期三 UTC+8下午8:57:23，Yuchao Jiang写道：

[Yuchao Jiang]

Bigger numchain and more importantly more sampling per chain.

To view this discussion on the web visit https://groups.google.com/d/msgid/canopy_phylogeny/105cc2bf-d744-4028-b5c7-c7531e64a46d%40googlegroups.com.

[ame...@b612.email]

Hi Yuchao,

Thanks for your quick reply.

I want to use the CAnopy to analyze my data, But I have several questions, the following are my questions.

1. I saw the website https://github.com/yuchaojiang/Canopy/blob/master/instruction/SNA_CNA_input.md, CNA copy number output by sequenza are A copy number and B copy number, and they are all integers. But the sequenza output A and B copy number are not the WM and Wm of Canopy. I use the Bf and depth ratio output by sequenza to calculate WM and Wm. Bf = Wm / (WM + Wm) and the depth ratio is depth.ratio = (WM + Wm)/2. That mean WM and Wm of Canopy are determined by BF and depth ratio output by sequenza, is that right?

2. I also encountered the error “Error in config.summary[i, 1] : subscript out of bounds”.  

    k=3:6, numchain=20, burin=10, thin=5, You recommend set thin and burin values according to the plot likelihood output of canopy.sample(), can you tell me how can I determine thin and burin? The attachment is my output plot likelihood figure.

3. What's more, I do not have multi-samples from the same patient, I just have a tumor and matched normal sample ( total is 2 samples), Is Canopy still fit for my situation?

sincerely,

Amelie
在 2018年11月15日星期四 UTC+8下午10:45:06，Yuchao Jiang写道：

[Yuchao Jiang]

On Nov 16, 2018, at 12:22 AM, ame...@b612.email wrote:

Hi Yuchao,

Thanks for your quick reply.

I want to use the CAnopy to analyze my data, But I have several questions, the following are my questions.

1. I saw the website https://github.com/yuchaojiang/Canopy/blob/master/instruction/SNA_CNA_input.md, CNA copy number output by sequenza are A copy number and B copy number, and they are all integers. But the sequenza output A and B copy number are not the WM and Wm of Canopy. I use the Bf and depth ratio output by sequenza to calculate WM and Wm. Bf = Wm / (WM + Wm) and the depth ratio is depth.ratio = (WM + Wm)/2. That mean WM and Wm of Canopy are determined by BF and depth ratio output by sequenza, is that right?

Yes although we recommend FALCON or FALCON-X to generate WM and Wm directly instead of transforming from log ratio and B allele frequency.

2. I also encountered the error “Error in config.summary[i, 1] : subscript out of bounds”.  

    k=3:6, numchain=20, burin=10, thin=5, You recommend set thin and burin values according to the plot likelihood output of canopy.sample(), can you tell me how can I determine thin and burin? The attachment is my output plot likelihood figure.

I’ve replied to a different thread. For future questions, you need to try to solve this by yourself first and aggregate your questions. We won’t have the bandwidth to reply to each of your email.

3. What's more, I do not have multi-samples from the same patient, I just have a tumor and matched normal sample ( total is 2 samples), Is Canopy still fit for my situation?

No. Not suited for one sample scenario.

To view this discussion on the web visit https://groups.google.com/d/msgid/canopy_phylogeny/16b5ce40-7bad-4e46-8219-0a57c68f1745%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

<toy_likelihood.pdf>