Error when running runDWLSDeconv

[martina-morchio]

Hi! I am using Giotto to analyse the transcriptome detected with Visium on 6 tissue sections, aiming to perform cell-type deconvolution with a single nuclei RNA seq dataset from two similar samples.

DWLS deconvolution works well when I'm analysing one tissue section individually, however to analyse all sections in parallel I joined the Giotto objects and performed Harmony integration.
All looks great until I run runDWLSDeconv, which starts for a while and then prints out the following message:

testcombo = runDWLSDeconv(gobject = testcombo,
                          sign_matrix = DWLS_matrix, cluster_column = "leiden_harmony")
Original error message: 

matrix D in quadratic function is not positive definite!
 Will try to fix error with Matrix::nearPD()
Error in Matrix::nearPD(D) : Matrix seems negative semi-definite
In addition: Warning message:
In asMethod(object) :
  sparse->dense coercion: allocating vector of size 1.3 GiB

I'm using the same DWLS matrix that worked for the individual sample.
Do you know what the issue might be? I have seen this error popping up in other issues (#121, #170, #228) but couldn't find a solution there.

Here's my session info:

> sessionInfo()
R version 4.3.1 (2023-06-16)
Platform: x86_64-conda-linux-gnu (64-bit)
Running under: CentOS Linux 7 (Core)

Matrix products: default
BLAS/LAPACK: /mnt/parscratch/users/mdq19mm/miniconda3/envs/giotto/lib/libopenblasp-r0.3.24.so;  LAPACK version 3.11.0

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

time zone: Europe/London
tzcode source: system (glibc)

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

other attached packages:
 [1] harmony_1.1.0               Rcpp_1.0.11                
 [3] scran_1.30.0                scuttle_1.12.0             
 [5] SingleCellExperiment_1.24.0 SummarizedExperiment_1.32.0
 [7] Biobase_2.62.0              GenomicRanges_1.54.0       
 [9] GenomeInfoDb_1.38.0         IRanges_2.36.0             
[11] S4Vectors_0.40.0            BiocGenerics_0.48.0        
[13] MatrixGenerics_1.14.0       matrixStats_1.0.0          
[15] ggplot2_3.4.4               Giotto_3.3.1               

loaded via a namespace (and not attached):
 [1] bitops_1.0-7              rlang_1.1.1              
 [3] magrittr_2.0.3            RcppAnnoy_0.0.21         
 [5] compiler_4.3.1            DelayedMatrixStats_1.24.0
 [7] png_0.1-8                 systemfonts_1.0.5        
 [9] vctrs_0.6.4               reshape2_1.4.4           
[11] RcppZiggurat_0.1.6        quadprog_1.5-8           
[13] stringr_1.5.0             pkgconfig_2.0.3          
[15] crayon_1.5.2              magick_2.8.1             
[17] XVector_0.42.0            labeling_0.4.3           
[19] utf8_1.2.4                ragg_1.2.6               
[21] purrr_1.0.2               Rfast_2.0.8              
[23] bluster_1.12.0            zlibbioc_1.48.0          
[25] beachmat_2.18.0           jsonlite_1.8.7           
[27] DelayedArray_0.28.0       tweenr_2.0.2             
[29] BiocParallel_1.36.0       terra_1.7-55             
[31] irlba_2.3.5.1             parallel_4.3.1           
[33] cluster_2.1.4             R6_2.5.1                 
[35] stringi_1.7.12            RColorBrewer_1.1-3       
[37] limma_3.58.0              reticulate_1.34.0        
[39] R.utils_2.12.2            FNN_1.1.3.2              
[41] Matrix_1.6-1.1            igraph_1.5.1             
[43] tidyselect_1.2.0          abind_1.4-5              
[45] codetools_0.2-19          lattice_0.22-5           
[47] tibble_3.2.1              plyr_1.8.9               
[49] withr_2.5.1               polyclip_1.10-6          
[51] scatterpie_0.2.1          pillar_1.9.0             
[53] ggfun_0.1.3               generics_0.1.3           
[55] dbscan_1.1-11             RCurl_1.98-1.12          
[57] sparseMatrixStats_1.14.0  munsell_0.5.0            
[59] scales_1.2.1              RhpcBLASctl_0.23-42      
[61] glue_1.6.2                metapod_1.10.0           
[63] tools_4.3.1               BiocNeighbors_1.20.0     
[65] data.table_1.14.8         ScaledMatrix_1.10.0      
[67] locfit_1.5-9.8            cowplot_1.1.1            
[69] grid_4.3.1                tidyr_1.3.0              
[71] edgeR_4.0.0               colorspace_2.1-0         
[73] GenomeInfoDbData_1.2.11   ggforce_0.4.1            
[75] BiocSingular_1.18.0       cli_3.6.1                
[77] rsvd_1.0.5                textshaping_0.3.7        
[79] fansi_1.0.5               S4Arrays_1.2.0           
[81] dplyr_1.1.3               uwot_0.1.16              
[83] gtable_0.3.4              R.methodsS3_1.8.2        
[85] digest_0.6.33             progressr_0.14.0         
[87] SparseArray_1.2.0         ggrepel_0.9.4            
[89] dqrng_0.3.1               farver_2.1.1             
[91] R.oo_1.25.0               lifecycle_1.0.3          
[93] statmod_1.5.0             MASS_7.3-60 
``` 

[martina-morchio]

As an update, I have run runDWLSDeconv() on all the individual samples that make up my larger dataset. It ran fine on most samples, but in 5 of them (out of 20) I encountered the same error message printed above.
I then merged and integrated the sections that ran successfully repeated runDWLSDeconv on this larger dataset, but again I encountered the matrix D error message, even though the deconvolution on those individual datasets worked fine when run individually.

[RubD]

@gcyuan could we take another look at this? I believe we have tried to catch errors, but not sure if this is a new one?