colGraph edges out-of-order when subsetting with un-ordered indeces

[nilseling]

Hi all,

I have been playing around with LoomExperiment and comparing the colGraphs functionality to SingleCellExperiment::colPairs.
I noticed a potential issue when subsetting a SingleCellLoomExperiment. Using the example file, I can see, that the 6th cell is connected to itself and the first cell (among others). The first cell is connected with the 6th cell but not itself (among others). Therefore scle[,c(1,6)] and scle[,c(6,1)] should return different edge lists in form of colGraphs entries. However:

library(LoomExperiment)
l1_file <-
  system.file("extdata", "L1_DRG_20_example.loom", package = "LoomExperiment")
scle <- import(l1_file, type="SingleCellLoomExperiment")
scle

ind <- c(1, 6)

scle2 <- scle[,ind]

colGraphs(scle2)[[1]]

returns

LoomGraph object with 3 hits and 1 metadata column:
           from        to |         w
      <integer> <integer> | <numeric>
  [1]         2         1 | 0.0714286
  [2]         1         2 | 0.0188679
  [3]         2         2 | 0.0406144
  -------
  nnode: 2

and

ind <- c(6, 1)

scle2 <- scle[,ind]

colGraphs(scle2)[[1]]

returns

LoomGraph object with 3 hits and 1 metadata column:
           from        to |         w
      <integer> <integer> | <numeric>
  [1]         2         1 | 0.0714286
  [2]         1         2 | 0.0188679
  [3]         2         2 | 0.0406144
  -------
  nnode: 2

When converting the SCLE into an SCE object, the subsetting operation works as expected:

library(SingleCellExperiment)
sce <- as(scle, "SingleCellExperiment")
colPair(sce, "KNN") <- as(colGraphs(scle)[[1]], "SelfHits")

ind <- c(6, 1)
sce2 <- sce[,ind]

colPair(sce2, "KNN")

which returns

SelfHits object with 3 hits and 1 metadata column:
           from        to |         w
      <integer> <integer> | <numeric>
  [1]         1         1 | 0.0406144
  [2]         1         2 | 0.0714286
  [3]         2         1 | 0.0188679
  -------
  nnode: 2

This is my session info:

R version 4.4.1 (2024-06-14)
Platform: x86_64-pc-linux-gnu
Running under: Red Hat Enterprise Linux 8.8 (Ootpa)

Matrix products: default
BLAS/LAPACK: FlexiBLAS OPENBLAS;  LAPACK version 3.11.0

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C               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    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       

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

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

other attached packages:
 [1] LoomExperiment_1.24.0       BiocIO_1.16.0               rhdf5_2.50.2                SingleCellExperiment_1.28.1
 [5] SummarizedExperiment_1.36.0 Biobase_2.66.0              GenomicRanges_1.58.0        GenomeInfoDb_1.42.3        
 [9] IRanges_2.40.1              MatrixGenerics_1.18.1       matrixStats_1.5.0           S4Vectors_0.44.0           
[13] BiocGenerics_0.52.0         workflowr_1.7.1            

loaded via a namespace (and not attached):
 [1] SparseArray_1.6.2       stringi_1.8.4           lattice_0.22-6          digest_0.6.37           magrittr_2.0.3         
 [6] evaluate_1.0.3          grid_4.4.1              fastmap_1.2.0           rprojroot_2.0.4         jsonlite_1.9.1         
[11] Matrix_1.7-2            processx_3.8.6          whisker_0.4.1           ps_1.9.0                promises_1.3.2         
[16] httr_1.4.7              UCSC.utils_1.2.0        HDF5Array_1.34.0        abind_1.4-8             cli_3.6.3              
[21] rlang_1.1.6             crayon_1.5.3            XVector_0.46.0          DelayedArray_0.32.0     yaml_2.3.10            
[26] S4Arrays_1.6.0          tools_4.4.1             Rhdf5lib_1.28.0         httpuv_1.6.15           GenomeInfoDbData_1.2.13
[31] R6_2.6.1                lifecycle_1.0.4         git2r_0.36.2            zlibbioc_1.52.0         stringr_1.5.1          
[36] fs_1.6.6                callr_3.7.6             later_1.4.2             glue_1.8.0              Rcpp_1.0.14            
[41] xfun_0.50               rhdf5filters_1.18.1     rstudioapi_0.17.1       knitr_1.49              htmltools_0.5.8.1      
[46] rmarkdown_2.29          compiler_4.4.1          getPass_0.2-4  

[LiNk-NY]

Hi @nilseling
See the reprex below. It seems like the issue is in loomSelectHits?

I am not very familiar with the package but the code does seem to use a sequence from lowest to highest seq_len(max(c(6, 1))) regardless of the order of the input..

suppressPackageStartupMessages(library(LoomExperiment))
l1_file <-
    system.file("extdata", "L1_DRG_20_example.loom", package = "LoomExperiment")
le <- import(l1_file)
cg <- colGraphs(le)
LoomExperiment:::loomSelectHits(cg$KNN, c(6, 1))
#> LoomGraph object with 3 hits and 1 metadata column:
#>            from        to |         w
#>       <integer> <integer> | <numeric>
#>   [1]         2         1 | 0.0714286
#>   [2]         1         2 | 0.0188679
#>   [3]         2         2 | 0.0406144
#>   -------
#>   nnode: 2
LoomExperiment:::loomSelectHits(cg$KNN, c(1, 6))
#> LoomGraph object with 3 hits and 1 metadata column:
#>            from        to |         w
#>       <integer> <integer> | <numeric>
#>   [1]         2         1 | 0.0714286
#>   [2]         1         2 | 0.0188679
#>   [3]         2         2 | 0.0406144
#>   -------
#>   nnode: 2

^{Created on 2025-04-23 with reprex v2.1.1}

Perhaps @mtmorgan would like to comment?
Thanks!