Running the test_data

[dannlbol]

Very interested in using this technique but having trouble running the test data supplied. Apologies if I'm missing something tried on UBUNTU, Mac same issue (giving details from Mac test here

Going by the example given in SNV_DA_doc.pdf... although I changed the -M file as it seems very much like it refers to the NSCLC data I run the following...

Rscript snvDA.R -M ./test_data/NSCLC_disease_outcome/NSCLC.SNVM.csv -A intronic_SNV_model -D 11 -U Disease_Free -H Relapse -G intronic -Z 3 -O -J -W 10 -V 10 -I 15 -T 4 -L 10 -B 1000 -E 40

Receiving error ...

Error in { : task 1 failed - "No control observation."
Calls: %dopar% ->
Execution halted

attempted to traceback issues in an R session seems that issue is to do with the get_predictions function and is an error thrown by the roc function having NaN in the class_hat object passed to the roc function...

Line 219 AUC = as.numeric(ci.auc(roc(class, class_hat)))

This in turn relates to ...

Line 206 unrounded.prediction <- (prediction$variates[i] - class.1.centroid) / (class.2.centroid - class.1.centroid).

...specifically the prediction object not having a valid entry in the $variates slot.

Apologies if I am doing something stupid but wonder if you could inform me on this issue.

This is the sessionInfo()

R version 3.3.1 (2016-06-21)
Platform: x86_64-apple-darwin13.4.0 (64-bit)
Running under: OS X 10.12 (Sierra)

locale:
[1] en_GB.UTF-8/en_GB.UTF-8/en_GB.UTF-8/C/en_GB.UTF-8/en_GB.UTF-8

attached base packages:
[1] parallel stats graphics grDevices utils datasets methods base

other attached packages:
[1] gplots_3.0.1 pROC_1.8 mixOmics_6.1.1 ggplot2_2.1.0 lattice_0.20-33 MASS_7.3-45
[7] doParallel_1.0.10 iterators_1.0.8 foreach_1.4.3 optparse_1.3.2

loaded via a namespace (and not attached):
[1] Rcpp_0.12.7 RColorBrewer_1.1-2 plyr_1.8.4 bitops_1.0-6 tools_3.3.1 digest_0.6.10
[7] jsonlite_1.1 tibble_1.2 gtable_0.2.0 igraph_1.0.1 shiny_0.14.1 DBI_0.5-1
[13] dplyr_0.5.0 stringr_1.1.0 knitr_1.14 caTools_1.17.1 htmlwidgets_0.7 gtools_3.5.0
[19] grid_3.3.1 getopt_1.20.0 ellipse_0.3-8 R6_2.2.0 rgl_0.96.0 gdata_2.17.0
[25] tidyr_0.6.0 reshape2_1.4.2 corpcor_1.6.8 magrittr_1.5 scales_0.4.0 codetools_0.2-14
[31] htmltools_0.3.5 assertthat_0.1 mime_0.5 colorspace_1.2-7 xtable_1.8-2 httpuv_1.3.3
[37] KernSmooth_2.23-15 stringi_1.1.2 munsell_0.4.3

[biomrpaul]

Hi Danibol,

I was able to recreate your error. It looks like this error is specific to this dataset, probably because it is smaller than the breast cancer dataset for which the software was optimized. I will be working to produce a solution to this soon.

In the mean time, I confirmed that using the breast cancer dataset did not result in errors. For ample, you can run a variation of the below command:
Rscript snvDA.R -M test_data/Breast_hormone_status/breast_Varley_et_al_SNVM.csv -A intron_model -D 42 -U ERpos -H TRneg -G intron -Z 7 -O -J -W 10 -V 10 -I 2 -T 24 -L 10 -B 400 -E 60 -N 2

Please let me know if you have any other questions!

Thanks,
Matt

[dannlbol]

OK works for me too. Will try my own set! As you say the smaller size set maybe throwing it off.

Thanks!

[biomrpaul]

Hey dannlbol,

The bug has now been fixed. With the newer imputation procedure, some predicted class labels were "NA" because of non-informative features (i.e. all samples had the same VAF or NA). These features have been removed prior to training. I re-ran the example run and it works fine!

Thanks,
Matt