Unexplained Errors

[Linda Sundermann]

Hi Yuchao,

when I used Canopy on simulated data, it exited with two unexplained
errors. I'd be happy if you could explain me what happened.

The first error occurred in the posterior evaluation, after BIC chose 2
as population number:
>Error in output.tree.Z %*% output.tree.P : non-conformable arguments
>Calls: canopy.post

The second error occurred in another run after I clustered the SNAs
(number of mutation clusters is 2) and started the sampling process:
>Sample in tree space with 2 subclones
>    Running chain 1 out of 3 ...
>Error in dimnames(x) <- dn :
>    length of 'dimnames' [2] not equal to array extent
>Calls: canopy.sample.cluster -> initialsna -> colnames<-

I'd be happy to provide more information if necessary.
Thank you for your help,
Linda

[Yuchao Jiang]

Can you send the rda files? Gene can you help take a look?

--
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_phylogeny+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/canopy_phylogeny/92ea2c10-7c71-68dc-85f3-986fc9356053%40gmail.com.
For more options, visit https://groups.google.com/d/optout.

[Linda Sundermann]

Sure.

The rda file for the first error is "lin_num_2.CN_noise_True.tree_index_0.4__sampchain.RData". It contains the chains after sampling, called "sampchain". The error occurs when I use the following function:

> post = canopy.post(sampchain = sampchain, projectname = projectname, K = K, numchain = numchain, burnin = burnin, thin = thin, optK = optK, post.config.cutoff = 0.05)

with the parameters
* K = 2:7
* numchain = 3
* burnin = 50
* thin = 1
* optK = 2
* projectname = "test"


The rda file for the second error is "lin_num_4_CN_noise_True_tree_index_0_6_.RData". It contains all necessary data to call the sampling function, where the error occurs:

> sampchain = canopy.sample.cluster(R=R, X=X, sna_cluster=sna_cluster, WM=WM, Wm=Wm, epsilonM=epsilonM, epsilonm=epsilonm, C=NULL, Y=Y, K=K, numchain=numchain, max.simrun=max.simrun, min.simrun=min.simrun, writeskip=writeskip, projectname=projectname, cell.line=FALSE, plot.likelihood=TRUE)


Thanks for your help,
best,
Linda

Am 9/5/18 um 3:35 PM schrieb Yuchao Jiang:

[Gene Urrutia]

Hi Linda,

Thanks for your interest in canopy and the Marathon pipeline.

The first error likely occurred because the simulated sampchain data you sent contains only 1 sample "sample1".  Canopy requires multiple spatially or temporally separated tumor samples with a matched normal sample.  This is how the phylogeny can be estimated.

The second error likely occurs for the same reason, that the data contains only 1 sample.

Please see a full description of inputs at https://github.com/yuchaojiang/MARATHON

Thanks,
Gene

To unsubscribe from this group and stop receiving emails from it, send an email to canopy_phyloge...@googlegroups.com.

[Linda Sundermann]

Hi Gene,

thanks for your quick answer.
I thought that the input to Canopy consists of VAFs of SNAs of one to multiple samples, together with the allele-specific coverage ratios between one to multiple tumor samples and a matched normal sample for CNAs. Did I understand you correctly, that Canopy also needs this normal sample as input? I'm not using the Marathon pipeline, only Canopy.
And I did run Canopy on other datasets before that only consisted of 1 sample and there the described errors did not occur. They seem to only occur if BIC chooses 2 as population number or if after clustering SNAs, 2 is chosen as the number of clusters.

Many thanks for your help,
Linda

Am 9/7/18 um 5:18 AM schrieb Gene Urrutia:

[Gene Urrutia]

Hi Linda,

You are correct:

"I thought that the input to Canopy consists of VAFs of SNAs of one to multiple samples, together with the allele-specific coverage ratios between one to multiple tumor samples and a matched normal sample for CNAs"

The simulated sampchain data you sent contains 1 sample.  It is not matched.  If you could describe the input upstream of this data, it might help me to understand better.  I.e. if the input was as described above, why only 1 sample was returned in the sampchain data.

Thanks,

Gene

[Linda Sundermann]

Hi Gene,

oh, by "allele-specific coverage ratios between one to multiple tumor samples and a matched normal sample for CNAs" I meant that the matched normal sample is needed to compute the ratios. Once the ratios are computed, I do not understand for what the normal sample is needed.

The simulated data I use only consists of 1 sample. It contains one region with CNAs and multiple SNAs. You can have a look at the data in the file "lin_num_4_CN_noise_True_tree_index_0_6_.rdata" (see previous mail). It contains all the input data like WM, Wm, epsilonM, epsilonm, Y, R and X. The dataset for which the other error occurs is similar to this one.

Thank you,
Linda

Am 9/7/18 um 3:49 PM schrieb Gene Urrutia:

[Yuchao Jiang]

Hi Linda,

I just wanted to add on top of what Gene suggested in his previous emails — Canopy is designed for deconvolution using multiple tumor samples that are spatially and/or temporarily separated from the same patient. If you only have one sample, we don’t think the deconvolution can be easily done. Canopy is not a suited method for this situation and you may resort to others such as ABSOLUTE or Sequenza.

Yuchao

[Linda Sundermann]

Hi Yuchao,

thanks for adding this and thank for suggesting alternative methods.

Best regards,
Linda

Am 9/9/18 um 11:07 PM schrieb Yuchao Jiang: