model optimization error bbdml and differentialTest

[amy-briggs]

Hello,

When I run differentialTest() and bbdml() using corncob v. 0.2.0, I keep getting the error, "Model could not be optimized! Try changing initializations or simplifying your model." I was able to fit these models successfully on this data in December, but have gotten a new computer and updated R from version 3.6 to 4.0.3 since then. When I attempted to run

corncob_da <- bbdml(formula = OTU.1 ~ DayAmdmt,
             phi.formula = ~ DayAmdmt,
             data = soil)

from the corncob vignette, I get the same error. I've tried reinstalling corncob and restarting R, without success. Do you have any other suggestions? Thank you in advance for any help.

Here is my Session Info:
[1] ‘0.2.0’

sessionInfo()
R version 4.0.3 (2020-10-10)
Platform: x86_64-apple-darwin17.0 (64-bit)
Running under: macOS Big Sur 10.16

Matrix products: default
LAPACK: /Library/Frameworks/R.framework/Versions/4.0/Resources/lib/libRlapack.dylib

attached base packages:
[1] grid stats4 stats graphics grDevices utils datasets methods base

other attached packages:
[1] magrittr_2.0.1 ggfortify_0.4.11 factoextra_1.0.7 glmmTMB_1.0.2.9000 scales_1.1.1
[6] egg_0.4.5 cowplot_1.1.1 gridExtra_2.3 corncob_0.2.0 microbiome_1.12.0
[11] phyloseq_1.34.0 data.table_1.13.6 bbmle_1.0.23.1 vegan_2.5-7 lattice_0.20-41
[16] permute_0.9-5 car_3.0-10 carData_3.0-4 lme4_1.1-26 Matrix_1.3-2
[21] forcats_0.5.0 stringr_1.4.0 purrr_0.3.4 readr_1.4.0 tidyr_1.1.2
[26] tibble_3.1.1 tidyverse_1.3.0 reshape2_1.4.4 dplyr_1.0.5 plyr_1.8.6
[31] MASS_7.3-53 ggplot2_3.3.3 dada2_1.18.0 Rcpp_1.0.5

loaded via a namespace (and not attached):
[1] estimability_1.3 coda_0.19-4 knitr_1.30
[4] DelayedArray_0.16.0 hwriter_1.3.2 RCurl_1.98-1.2
[7] generics_0.1.0 BiocGenerics_0.36.0 xml2_1.3.2
[10] lubridate_1.7.9.2 SummarizedExperiment_1.20.0 assertthat_0.2.1
[13] xfun_0.20 hms_1.0.0 evaluate_0.14
[16] fansi_0.4.2 progress_1.2.2 dbplyr_2.0.0
[19] readxl_1.3.1 igraph_1.2.6 DBI_1.1.1
[22] ellipsis_0.3.1 ggpubr_0.4.0 backports_1.2.1
[25] ROI_1.0-0 RcppParallel_5.0.2 MatrixGenerics_1.2.0
[28] vctrs_0.3.7 Biobase_2.50.0 abind_1.4-5
[31] withr_2.4.2 bdsmatrix_1.3-4 emmeans_1.5.3
[34] GenomicAlignments_1.26.0 prettyunits_1.1.1 cluster_2.1.0
[37] ape_5.4-1 crayon_1.4.1 slam_0.1-48
[40] pkgconfig_2.0.3 labeling_0.4.2 GenomeInfoDb_1.26.2
[43] nlme_3.1-151 rlang_0.4.10 lifecycle_1.0.0
[46] registry_0.5-1 modelr_0.1.8 cellranger_1.1.0
[49] matrixStats_0.57.0 Rhdf5lib_1.12.0 boot_1.3-25
[52] reprex_0.3.0 png_0.1-7 bitops_1.0-6
[55] ROI.plugin.lpsolve_1.0-0 rhdf5filters_1.2.0 Biostrings_2.58.0
[58] ShortRead_1.48.0 jpeg_0.1-8.1 rstatix_0.7.0
[61] S4Vectors_0.28.1 ggsignif_0.6.1 zlibbioc_1.36.0
[64] compiler_4.0.3 RColorBrewer_1.1-2 detectseparation_0.2
[67] Rsamtools_2.6.0 cli_2.5.0 ade4_1.7-16
[70] XVector_0.30.0 TMB_1.7.18 mgcv_1.8-33
[73] tidyselect_1.1.0 stringi_1.5.3 yaml_2.2.1
[76] latticeExtra_0.6-29 ggrepel_0.9.0 tools_4.0.3
[79] parallel_4.0.3 rio_0.5.16 rstudioapi_0.13
[82] foreach_1.5.1 foreign_0.8-81 farver_2.1.0
[85] Rtsne_0.15 digest_0.6.27 GenomicRanges_1.42.0
[88] broom_0.7.6 httr_1.4.2 colorspace_2.0-0
[91] rvest_0.3.6 fs_1.5.0 IRanges_2.24.1
[94] splines_4.0.3 statmod_1.4.35 multtest_2.46.0
[97] xtable_1.8-4 jsonlite_1.7.2 nloptr_1.2.2.2
[100] lpSolveAPI_5.5.2.0-17.7 R6_2.5.0 pillar_1.6.0
[103] htmltools_0.5.0 glue_1.4.2 minqa_1.2.4
[106] BiocParallel_1.24.1 codetools_0.2-18 mvtnorm_1.1-1
[109] utf8_1.2.1 numDeriv_2016.8-1.1 curl_4.3
[112] zip_2.1.1 openxlsx_4.2.3 survival_3.2-7
[115] rmarkdown_2.6 biomformat_1.18.0 munsell_0.5.0
[118] rhdf5_2.34.0 GenomeInfoDbData_1.2.4 iterators_1.0.13
[121] haven_2.3.1 gtable_0.3.0

[bryandmartin]

Hi @amy-briggs,

Thanks for posting this. I haven't been able to reproduce it despite having the same operating system and package info, so I might need to ask some follow up questions to figure out what's going on here.

This issue comes from an error in optimization. Can you try manually (re-)installing the "trust" and "optimr" packages? I'm wondering if they aren't loading properly for some reason.

Bryan

[amy-briggs]

Oh, that worked! I'm so glad it was that simple. Thank you!