Crash in the qualification workflow

[Yuri05]

Running this workflow:

i Info	Starting run of Plot Comparison Time Profiles task
05/09/2024 - 16:53:57
! Warning	Problem while converting geom to grob.
	
> Error Trace
	 ↪ createQualificationReport(qualificationRunnerFolder = "C:/QualificationRunner/", 
	  ↪ workflow$runWorkflow()
	   ↪ self[[plotTask]]$runTask(self$configurationPlan)
	    ↪ self$saveResults(taskResults, configurationPlan)
	     ↪ result$saveFigure(fileName = figureFilePath)
	      ↪ updatePlotDimensions(self$plot)
	       ↪ ggplot2::ggplotGrob(plotObject)
	        ↪ ggplot_gtable(ggplot_build(x))
	         ↪ ggplot_gtable.ggplot_built(ggplot_build(x))
	          ↪ by_layer(function(l, d) l$draw_geom(d, layout), plot$layers, 
	           ↪ try_fetch(for (i in seq_along(data)) {
	            ↪ f(l = layers[[i]], d = data[[i]])
	             ↪ l$draw_geom(d, layout)
	              ↪ draw_geom(..., self = self)
	               ↪ self$geom$draw_layer(data, self$computed_geom_params, layout, 
	                ↪ draw_layer(..., self = self)
	                 ↪ lapply(data_panels, function(data) {
	                  ↪ FUN(X[[i]], ...)
	                   ↪ inject(self$draw_panel(data, panel_params, coord, !!!params))
	                    ↪ self$draw_panel(data, panel_params, coord, na.rm = TRUE, outline.type = "both", 
	                     ↪ draw_panel(..., self = self)
	                      ↪ lapply(groups, function(group) {
	                       ↪ FUN(X[[i]], ...)
	                        ↪ self$draw_group(group, panel_params, coord, ...)
	                         ↪ draw_group(..., self = self)
	                          ↪ polygonGrob(munched_poly$x, munched_poly$y, id = munched_poly$id, 
	                           ↪ grob(x = x, y = y, id = id, id.lengths = id.lengths, name = name, 
	                            ↪ gpar(fill = fill_alpha(aes$fill, aes$alpha), col = if (is_full_outline) aes$colour else NA, 
	                             ↪ validGP(list(...))
	                              ↪ check.length("fill")
	                               ↪ handlers[[1L]](cnd)
	                                ↪ cli::cli_abort(c("Problem while {step}.", i = "Error occurred in the {ordinal(i)} layer."), 
	                                 ↪ rlang::abort(message, ..., call = call, use_cli_format = TRUE, 
	                                  ↪ signal_abort(cnd, .file)
	                                   ↪ signalCondition(cnd)
	                                    ↪ (function (errorCondition) 
Error: Problem while converting geom to grob.
In addition: Warning message:
In readLines(fileName, encoding = "UTF-8") :
 
 Error: Problem while converting geom to grob. 

[Yuri05]

@pchelle we should fix it before release

[pchelle]

Error message looks like a {ggplot2} issue.
Could you provide the sessionInfo ?

[Yuri05]

> sessionInfo()
R version 4.2.2 (2022-10-31 ucrt)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 10 x64 (build 19045)

Matrix products: default

locale:
[1] LC_COLLATE=German_Germany.utf8  LC_CTYPE=German_Germany.utf8    LC_MONETARY=German_Germany.utf8
[4] LC_NUMERIC=C                    LC_TIME=German_Germany.utf8    

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

other attached packages:
[1] ospsuite.reportingengine_2.2.363 tlf_1.5.170                      ospsuite_12.1.0.9003            
[4] rSharp_1.0.0                    

loaded via a namespace (and not attached):
 [1] Rcpp_1.0.11           pillar_1.9.0          compiler_4.2.2        tools_4.2.2           sysfonts_0.8.8       
 [6] jsonlite_1.8.8        lifecycle_1.0.4       tibble_3.2.1          gtable_0.3.1          pkgconfig_2.0.3      
[11] rlang_1.1.3           cli_3.6.2             rstudioapi_0.15.0     commonmark_1.9.0      parallel_4.2.2       
[16] xfun_0.41             knitr_1.42            withr_3.0.0           dplyr_1.1.4           showtextdb_3.0       
[21] stringr_1.5.1         xml2_1.3.3            systemfonts_1.0.4     ospsuite.utils_1.5.35 generics_0.1.3       
[26] vctrs_0.6.5           gridtext_0.1.5        ggtext_0.1.2          grid_4.2.2            tidyselect_1.2.0     
[31] glue_1.7.0            data.table_1.14.8     R6_2.5.1              textshaping_0.3.6     fansi_1.0.4          
[36] farver_2.1.1          tidyr_1.3.0           ggplot2_3.5.1         purrr_1.0.2           magrittr_2.0.3       
[41] scales_1.3.0          showtext_0.9-6        colorspace_2.1-0      ragg_1.2.5            labeling_0.4.2       
[46] utf8_1.2.3            stringi_1.7.12        munsell_0.5.0         markdown_1.12         crayon_1.5.2         

[pchelle]

@Yuri05
I got a different crash before being able to plot anything because the simulations did not run through :(

Error messages

[CVODES WARNING]  CVode
  Internal t = 36243 and h = 1e-12 are such that t + h = t on the next step. The solver will continue anyway.

06/09/2024 - 03:51:41
! Warning	simpleWarning in .getConcurrentSimulationRunnerResults(results = results, resultsIdSimulationIdMap = resultsIdSimulationIdMap, : Simulation run failed: some variables became negative when trying to reach t=38322. There are different possible reasons for this:

  - Solver tolerances are too high. Please reduce the absolute and relative tolerances by one order of magnitude (e.g. from 1E-9 to 1E-10) and restart the simulation.
  - Some variables which are allowed to be negative were defined as non-negative.
  - Model is not properly established (e.g. the kinetic should be k*[A] but was defined as k*[B] etc.)

The following variables became negative :
Schoedel (2012) - DDI - quinidine-dextromethorphan-paroxetine, 30/30/20 mg, po, n=14 (EM)|Organism|Gonads|BloodCells|Dextromethorphan-CYP2D6-Lutz 2012 (dextrorphan) Metabolite

06/09/2024 - 03:51:41
! Warning	Simulation of model file 'Crash_DDGI/re_input/Quinidine-Dextromethorphan-DDI/Capon (1996) - dextromethorphan hydrobromide, 30 mg, po, n=6 (EM)//Capon (1996) - dextromethorphan hydrobromide, 30 mg, po, n=6 (EM).pkml' from simulation set 'Quinidine-Dextromethorphan-DDI-Capon (1996) - dextromethorphan hydrobromide, 30 mg, po, n=6 (EM)' was not completed successfully.; ...

I added ... but the error message included many simulations if not all

By any chance, could you share the simulation results of the first comparison plot ? Or is there a setting I can update to prevent such issue ?

[Yuri05]

This is very strange. Which version of ospsuite package are you using? Mine is 12.1.0.9003
Here are simulation results and PK-Analysis calculation results of the workflow. You can extract them into the re_output folder of the workflow, deactivate the corresponding tasks and try to rerun again.
re_output.zip

[pchelle]

This is very strange. Which version of ospsuite package are you using? Mine is 12.1.0.9003

Same: ospsuite_12.1.0.9003
All my versions seem aligned except for rSharp, would rSharp_1.0.0.9000 be the reason then ?

[Yuri05]

All my versions seem aligned except for rSharp, would rSharp_1.0.0.9000 be the reason then ?

no. it just propagates inputs/outputs to the "working" components embedded within the ospsuite package.

[pchelle]

I was able to find the cause of the crash:
Shadel 1995 simulated data in second comparison time profile plot has only 0 values which makes the log scale plot somehow.
The culprit seems to be the behaviour of {ggplot2} from the oob argument in setYAxis.

Not sure if the {tlf} issue has always been there or not since I feel that the simulated concentrations should not be 0 here.

[Yuri05]

Not sure if the {tlf} issue has always been there or not since I feel that the simulated concentrations should not be 0 here.

Probably the issue was always there.
Theoretically, sim concentration should not be const zero, but if it's the case: log plot should be just empty and not crash.

[pchelle]

Indeed the issue was always there and comes from {ggplot2} handling of legends
https://github.com/Open-Systems-Pharmacology/TLF-Library/blob/2257c8f677e700382e99937769fffd59a6beb17e/R/utilities-axis.R#L303-L312

Censoring Infinite values from the log is appropriate but crashes when used on ribbons.
However, {tlf} has been adding layers of point, line and ribbon to enable a convenient way of displaying a unique legend in time profiles.