Train marker-free classifier: Error in .testForValidKeys(x, keys, keytype, fks)

[Moritz Przybilla]

Hey Hannah et al.,

I am trying to use Garnett's marker free classification, for the classification of some stomach tissue cells, but unfortunately encounter an error when following your vignette. I use the following code to run convert my seurat object to the monocle_cds:

seurat.obj.big <- readRDS("my_seurat_object")

### Construct the cds object

#Extract data, phenotype data, and feature data from the SeuratObject
data <- as(as.matrix(seurat.obj.big@assays$RNA@counts), 'sparseMatrix')

pData <- seurat.obj.big@meta.data
fData <- data.frame(gene_short_name = row.names(data), row.names = row.names(data))

#Construct monocle cds
monocle_cds <- monocle3::new_cell_data_set(expression_data = data,
                                           cell_metadata = pData,
                                           gene_metadata = fData)

# provide cluster_ids to garnett
pData(monocle_cds)$garnett_cluster <- pData(monocle_cds)$seurat_clusters

My marker file looks like this, where "CellType" corresponds to a column in the metadata.

>PMCs
CellType: PMCs
references: https://www.cell.com/cell-reports/pdf/S2211-1247(19)30525-X.pdf

>GMCs
CellType: GMCs

>MSCs
CellType: MSCs

>Enterocytes
CellType: Enterocytes

>Enteroendocrine
CellType: Enteroendocrine

>PCs
CellType: PCs

>Cancer Cells
expressed: Cancer Cells

>Goblet Cells
CellType: Goblet Cells

>Neck-like Cells
CellType: Neck-like Cells

>Chief Cells
CellType: Chief Cells

>Endocrine Cells
CellType: Endocrine Cells

>Parietal Cells
CellType: Parietal Cells

If I then want to train my classifier I get the following error:

# train classifier
gastric_classifier <- train_cell_classifier(cds = monocle_cds,
                                            marker_file = garnett.marker.file,
                                            db=org.Hs.eg.db,
                                            cds_gene_id_type = "SYMBOL",
                                            num_unknown = 50,
                                            marker_file_gene_id_type = "SYMBOL")

There are 12 cell type definitions
Error in value[[3L]](cond) : Garnett cannot convert the gene IDs using the db and types provided. Please check that your db, cds_gene_id_type and marker_file_gene_id_type parameters are correct. Please note that the  
cds_gene_id_type refers to the type of the row.names of the feature (gene) table in your cds. Conversion error: Error in .testForValidKeys(x, keys, keytype, fks): None of the keys entered are valid keys for 'SYMBOL'. Please use the keys method to see a listing of valid arguments.

I used the exact same cds object for classification with the prior knowledge (marker-gene) classifier and it worked without any problems...so I do not really know, whats going wrong..

For completeness rownames and matrix look like this:

> head(fData(monocle_cds))
DataFrame with 6 rows and 1 column
        gene_short_name
               <factor>
STPG1             STPG1
NIPAL3           NIPAL3
AK2                 AK2
KDM1A             KDM1A
DNAJC11         DNAJC11
E2F2               E2F2

> head(exprs(monocle_cds)[,1:10])
6 x 10 sparse Matrix of class "dgCMatrix"
   [[ suppressing 10 column names ‘CAG1_AAACCTGCACAGGTTT-1’, ‘CAG1_AAACCTGCATCCTTGC-1’, ‘CAG1_AAACGGGGTCTTCGTC-1’ ... ]]
                           
STPG1   . . . . . . . . . .
NIPAL3  . . . . . . . . . .
AK2     1 . . . . . . . . .
KDM1A   1 . . . . . 1 . . .
DNAJC11 . . . . . . . . . .
E2F2    . . . . . . . . . .

Would greatly appreciate your help here :) Thanks a lot!

Very best wishes,

Moritz

> sessionInfo()
R version 3.6.0 (2019-04-26)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: CentOS Linux 7 (Core)

Matrix products: default
BLAS/LAPACK: /usr/lib64/libopenblas-r0.3.3.so

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C               LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8     LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8    LC_PAPER=en_US.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C             LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

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

other attached packages:
 [1] garnett_0.2.11              monocle3_0.2.1.9            SingleCellExperiment_1.8.0  org.Hs.eg.db_3.10.0         AnnotationDbi_1.48.0        DESeq2_1.26.0               SummarizedExperiment_1.16.1 DelayedArray_0.12.3        
 [9] BiocParallel_1.20.1         matrixStats_0.56.0          Biobase_2.46.0              GenomicRanges_1.38.0        GenomeInfoDb_1.22.1         IRanges_2.20.2              S4Vectors_0.24.4            BiocGenerics_0.32.0        
[17] cowplot_1.0.0               patchwork_1.0.0             forcats_0.5.0               stringr_1.4.0               purrr_0.3.4                 readr_1.3.1                 tidyr_1.0.2                 tibble_3.0.1               
[25] tidyverse_1.3.0             umap_0.2.5.0                reticulate_1.15             Rtsne_0.15                  gridGraphics_0.5-0          beanplot_1.2                RColorBrewer_1.1-2          scales_1.1.0               
[33] viridis_0.5.1               viridisLite_0.3.0           ggridges_0.5.2              ggplot2_3.3.0               svd_0.5                     qlcMatrix_0.9.7             sparsesvd_0.2               slam_0.1-47                
[41] Matrix_1.2-18               dplyr_0.8.5                 Seurat_3.1.5                devtools_2.3.0              usethis_1.6.1               robustbase_0.93-6          

loaded via a namespace (and not attached):
  [1] tidyselect_1.0.0       RSQLite_2.2.0          htmlwidgets_1.5.1      docopt_0.6.1           munsell_0.5.0          codetools_0.2-16       ica_1.0-2              future_1.17.0          withr_2.2.0            colorspace_1.4-1      
 [11] knitr_1.28             rstudioapi_0.11        ROCR_1.0-7             listenv_0.8.0          GenomeInfoDbData_1.2.2 bit64_0.9-7            rhdf5_2.30.1           rprojroot_1.3-2        vctrs_0.3.0            generics_0.0.2        
 [21] lambda.r_1.2.4         xfun_0.13              R6_2.4.1               clue_0.3-57            rsvd_1.0.3             locfit_1.5-9.4         bitops_1.0-6           assertthat_0.2.1       nnet_7.3-14            gtable_0.3.0          
 [31] npsurv_0.4-0           globals_0.12.5         processx_3.4.2         rlang_0.4.6            genefilter_1.68.0      GlobalOptions_0.1.1    splines_3.6.0          lazyeval_0.2.2         acepack_1.4.1          checkmate_2.0.0       
 [41] broom_0.5.6            BiocManager_1.30.10    reshape2_1.4.4         modelr_0.1.7           backports_1.1.6        Hmisc_4.4-0            tools_3.6.0            ellipsis_0.3.0         gplots_3.0.3           sessioninfo_1.1.1     
 [51] Rcpp_1.0.4.6           plyr_1.8.6             base64enc_0.1-3        zlibbioc_1.32.0        RCurl_1.98-1.2         ps_1.3.2               prettyunits_1.1.1      rpart_4.1-15           openssl_1.4.1          pbapply_1.4-2         
 [61] GetoptLong_0.1.8       zoo_1.8-8              haven_2.2.0            ggrepel_0.8.2          cluster_2.1.0          fs_1.4.1               magrittr_1.5           futile.options_1.0.1   data.table_1.12.8      RSpectra_0.16-0       
 [71] circlize_0.4.9         reprex_0.3.0           lmtest_0.9-37          RANN_2.6.1             packrat_0.5.0          fitdistrplus_1.0-14    pkgload_1.0.2          xtable_1.8-4           hms_0.5.3              lsei_1.2-0            
 [81] XML_3.99-0.3           jpeg_0.1-8.1           readxl_1.3.1           gridExtra_2.3          shape_1.4.4            testthat_2.3.2         compiler_3.6.0         KernSmooth_2.23-17     crayon_1.3.4           htmltools_0.4.0       
 [91] Formula_1.2-3          geneplotter_1.64.0     lubridate_1.7.8        DBI_1.1.0              formatR_1.7            dbplyr_1.4.3           ComplexHeatmap_2.2.0   MASS_7.3-51.6          rappdirs_0.3.1         cli_2.0.2             
[101] gdata_2.18.0           igraph_1.2.5           pkgconfig_2.0.3        foreign_0.8-76         plotly_4.9.2.1         xml2_1.3.2             annotate_1.64.0        ArchR_0.9.3            XVector_0.26.0         rvest_0.3.5           
[111] callr_3.4.3            digest_0.6.25          sctransform_0.2.1      RcppAnnoy_0.0.16       tsne_0.1-3             cellranger_1.1.0       leiden_0.3.3           htmlTable_1.13.3       uwot_0.1.8             curl_4.3              
[121] gtools_3.8.2           rjson_0.2.20           lifecycle_0.2.0        nlme_3.1-147           jsonlite_1.6.1         Rhdf5lib_1.8.0         futile.logger_1.4.3    desc_1.2.0             askpass_1.1            fansi_0.4.1           
[131] pillar_1.4.3           lattice_0.20-41        ggrastr_0.1.7          httr_1.4.1             DEoptimR_1.0-8         pkgbuild_1.0.7         survival_3.1-12        glue_1.4.0             remotes_2.1.1          rly_1.6.2             
[141] png_0.1-7              bit_1.1-15.2           stringi_1.4.6          blob_1.2.1             latticeExtra_0.6-29    caTools_1.18.0         memoise_1.1.0          irlba_2.3.3            future.apply_1.5.0     ape_5.3          

[Hannah Pliner]

Hi Moritz,

Looks to me like the issue is with the 'Cancer Cells' definition - looks like you forgot to swap out 'expressed' for 'CellType'. When the code sees that none of your 'marker genes' (in this case 'Cancer Cells') are in the database, it assumes something went wrong! 

Let me know if that wasn't the fix!

Hannah

[Moritz Przybilla]

Oh, yes - absolutely! Such an easy mistake - totally missed it!

Yes, that fixed it. Thank you so much for taking a look - appreciate it!

Best,

Moritz

[Sahni Lab]

Hi everyone,

I hope it is ok that I keep using this very old thread, but I encountered the same error and cannot find my error in the file as was the solution for this example.

I have successfully tested the marker-bound classifier by using this code:

` bs_classifier_free <- train_cell_classifier(cds = bs_cds,

                                         marker_file =  "D:\\Sahni\\Julia\\scSeq\\Garnett\\Garnett_top10_pos.txt",
                                         db=org.Mm.eg.db,

                                         cds_gene_id_type = "SYMBOL",
                                         num_unknown = 50,

                                         marker_file_gene_id_type = "SYMBOL") `

 

result>

` There are 6 cell type definitions training_sample B.lat B.med S.lat S.med S.med.1 S.med.2 Unknown 500 500 500 500 500 500 50 Model training finished.`

However, when i link to the marker-free file, which looks like this (here, I removed the dots in the names as I wasnt sure if this would cause the error, but they are also removed in the corresponding column):

` >Slat
clusters: Slat

>Smed
clusters: Smed

>Bmed
clusters: Bmed

>Blat
clusters: Blat

>Smed1
clusters: Smed1

>Smed2
clusters: Smed2 `

(and clusters is a column in the bs_cds), I get this error:

Error in value[[3L]](cond) : Garnett cannot convert the gene IDs using the db and types provided. Please check that your db, cds_gene_id_type and marker_file_gene_id_type parameters are correct. Please note that the cds_gene_id_type refers to the type of the row.names of the feature (gene) table in your cds. Conversion error: Error in mapIds_base(x, keys, column, keytype, ..., multiVals = multiVals): mapIds must have at least one key to match against.

Is there a way to troubleshoot if it actually is able to find the cluster column in the CDS file?

Thanks so much for any help you can provide!

[hpl...@gmail.com]

Sorry for the very long delay.

I don't see an obvious issue here... you're certain that the column name is 'clusters' instead of 'cluster'?  Can you send the output of traceback() after the error to see exactly where the issue is?