Search
Question: 'L-BFGS-B needs finite values of 'fn''-error when running Clomial package
0
gravatar for Jouni Kujala
6 days ago by
Finland
Jouni Kujala0 wrote:

Hi!

I have tried to test Clomial R package to infer the clonal structure of panel sequenced tumor samples but I am facing some problems when trying to run my samples with commands. I have tried to debug it but so far I have not been able to solve this. I believe that reason for this odd behavior is related to my low variant coverage (see example below). To confirm this, I am kindly requesting some technical support to help me overcome this problem. Package seems to be easy to use and I would be more than happy to utilize it in my research project.

> library(Clomial)
> set.seed(1)

> Dc = read.table(“sample_Dc.txt”, header=T, row.names=1)
> head(Dc)
                 sampleA   sampleB
chr1:27056288         86    415
chr1:27057762        507    611
chr1:27088785         97    277
chr2:178107590       174    673
chr2:198257616         9     55
chr2:198267216       174    289

> Dt = read.table(“sample_Dt.txt”, header=T, row.names=1)

> head(Dt)
                 sampleA   sampleB
chr1:27056288         12         0
chr1:27057762         45         0
chr1:27088785          6         0 
chr2:178107590        16         0
chr2:198257616         0         8
chr2:198267216        18         0


> clomialResults = Clomial(Dc=Dc, Dt=Dt, maxIt=4, C=4, binomTryNum=2)

Error in optim(par = Wj, fn = objective, gr = gradient, lower = lowers, :
      L-BFGS-B need finite values of ‘fn’
Error in optim(par = Wj, fn = objective, gr = gradient, lower = lowers, :
      L-BFGS-B need finite values of ‘fn’
Error in optim(par = Wj, fn = objective, gr = gradient, lower = lowers, :
      L-BFGS-B need finite values of ‘fn’
Error in optim(par = Wj, fn = objective, gr = gradient, lower = lowers, :
      L-BFGS-B need finite values of ‘fn’
Error in optim(par = Wj, fn = objective, gr = gradient, lower = lowers, :
      L-BFGS-B need finite values of ‘fn’


Warning messages;
1:  In log(row.z.likelihoods[i, state]) : NaNs produced
2:  In log(row.z.likelihoods[i, state]) : NaNs produced
3:  In log(row.z.likelihoods[i, state]) : NaNs produced
4:  In log(row.z.likelihoods[i, state]) : NaNs produced
5:  In log(row.z.likelihoods[i, state]) : NaNs produced

> systemInfo()

R version 3.2.2 (2015-08-14)
Platform: x86_64-redhat-linux-gnu (64-bit)
Running under: CentOS release 6.9 (Final)

locale:
[1] LC_TYPE=en_US.UTF-8
[2] LC_NUMERIC=C
[3] LC_TIME=en_US:UTF-8
[4] LC_COLLATE=en_US:UTF-8
[5] LC_MONETARY=en_US:UTF-8
[6] LC_MESSAGES=en_US:UTF-8
[7] LC_PAPER=en_US:UTF-8
[8] LC_NAME=C
[9] LC_ADDRESS=C
[10] LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_US:UTF-8
[12] LC_IDENTIFICATION=C

attached base packages:
[1] stats
[2] graphics
[3] qrDevices utils
[4] datasets
[5] methods
[6] base

other attached packages:
[1] Clomial_1.6.0       matrixStats_0.52.2

loaded via a namespace (and not attached):
[1] tools_3.2.2

 

Best regards,

Jouni

ADD COMMENTlink modified 5 days ago by Habil Zare60 • written 6 days ago by Jouni Kujala0
0
gravatar for Habil Zare
5 days ago by
Habil Zare60
United States/Austin Area/Texas State University
Habil Zare60 wrote:

1- Two samples are not enough to infer the clonal composition.

2- There are 4 samples in the data you sent me by email. The mutation ratios seem to be very low (~0.01).

plot(rowSums(Dc), rowSums(Dt),xlim=c(0,1000), ylim=c(0,30))

  

Your data has 169 mutations, and with only 4 samples, it's not easy to fit the binomial model to the data. The system is underconditioned. I could run Clomial using 50 random mutations with no error:

ClomialRes50 <- Clomial(Dc=Dc[1:50,], Dt=Dt[1:50,], maxIt=4, C=4, binomTryNum=2)

I recommend increasing the number of samples if possible. If not, filter the mutations. Say, include those mutations with a frequency of at least 2% in at least one of the samples.

 

ADD COMMENTlink written 5 days ago by Habil Zare60

Thank you Habil for your response and a thorough explanation! I have tried Clomial with a few sample sets where the number of samples varies between 2 and 5 samples per patient case. I now see that the number of samples is a major problem in my case and I'm afraid that I am not able to increase the number of samples due to nature of our study.

I will keep trying to run Clomial with sample sets where the number of samples is highest. I also agree that the mutation ratio (allele frequency) of my mutation data is very low and we will try to filter the rarest mutations as you suggested and try to run Clomial with this input.

 

ADD REPLYlink written 5 days ago by Jouni Kujala0

You're welcome! I am interested in Clomial performance if you will have a way to assess it.

ADD REPLYlink written 5 days ago by Habil Zare60
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.2.0
Traffic: 417 users visited in the last hour