Cannot create idx.annoy for mouse pansci reference Azimuth

[ruggerobelluomo]

Hi all,
I downloaded the mouse pansci reference object from here: https://zenodo.org/records/14042645
However, the annoy index is not present at this link so I tried to create it myself following a previous issue:
obj.azimuth=readRDS("C:/Users/rugge/Downloads/mouse_pansci/pansci_filtered.rds")
SaveAnnoyIndex(object = obj.azimuth[["refdr.annoy.neighbors"]], file = "C:/Users/rugge/Downloads/mouse_pansci//idx.annoy")
And got this error:
Error in obj.azimuth[["refdr.annoy.neighbors"]]:
! 'refdr.annoy.neighbors' not found in this Seurat object

Run rlang::last_trace() to see where the error occurred

Any help please? Thanks!

sessionInfo()
R version 4.3.2 (2023-10-31 ucrt)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 11 x64 (build 22631)

Matrix products: default

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

time zone: Europe/Amsterdam
tzcode source: internal

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

other attached packages:
[1] SeuratData_0.2.2.9001 shiny_1.8.0 org.Mm.eg.db_3.17.0 AnnotationDbi_1.62.2
[5] harmony_1.2.0 Rcpp_1.0.11 clusterProfiler_4.8.3 SingleR_2.2.0
[9] scater_1.28.0 scran_1.28.2 scuttle_1.10.3 data.table_1.15.2
[13] reticulate_1.35.0 SingleCellExperiment_1.22.0 SummarizedExperiment_1.30.2 Biobase_2.60.0
[17] GenomicRanges_1.52.0 GenomeInfoDb_1.36.4 IRanges_2.34.1 S4Vectors_0.38.1
[21] BiocGenerics_0.46.0 MatrixGenerics_1.12.3 matrixStats_1.2.0 sctransform_0.4.1
[25] pheatmap_1.0.12 SeuratWrappers_0.3.1 cowplot_1.1.3 dplyr_1.1.4
[29] DoubletFinder_2.0.4 EnhancedVolcano_1.18.0 ggrepel_0.9.5 ggplot2_3.5.0
[33] Seurat_5.1.0 SeuratObject_5.0.2 sp_2.1-2 Azimuth_0.5.0
[37] shinyBS_0.61.1

loaded via a namespace (and not attached):
[1] R.methodsS3_1.8.2 progress_1.2.3 poweRlaw_0.80.0
[4] goftest_1.2-3 DT_0.32 Biostrings_2.68.1
[7] vctrs_0.6.5 spatstat.random_3.2-3 digest_0.6.33
[10] png_0.1-8 deldir_2.0-4 parallelly_1.37.0
[13] MASS_7.3-60 Signac_1.14.0 reshape2_1.4.4
[16] qvalue_2.32.0 httpuv_1.6.14 withr_3.0.2
[19] ggfun_0.1.4 ellipsis_0.3.2 survival_3.5-7
[22] EnsDb.Hsapiens.v86_2.99.0 memoise_2.0.1 ggbeeswarm_0.7.2
[25] gson_0.1.0 tidytree_0.4.6 zoo_1.8-12
[28] gtools_3.9.5 pbapply_1.7-2 R.oo_1.26.0
[31] prettyunits_1.2.0 KEGGREST_1.40.1 promises_1.2.1
[34] httr_1.4.7 downloader_0.4 restfulr_0.0.15
[37] globals_0.16.3 fitdistrplus_1.1-11 rhdf5filters_1.12.1
[40] ps_1.7.6 rhdf5_2.44.0 rstudioapi_0.15.0
[43] DOSE_3.26.2 miniUI_0.1.1.1 generics_0.1.3
[46] processx_3.8.4 curl_5.2.1 zlibbioc_1.46.0
[49] ggraph_2.2.1 ScaledMatrix_1.8.1 polyclip_1.10-6
[52] GenomeInfoDbData_1.2.10 xtable_1.8-4 stringr_1.5.1
[55] desc_1.4.3 pracma_2.4.4 S4Arrays_1.0.6
[58] BiocFileCache_2.8.0 hms_1.1.3 irlba_2.3.5.1
[61] colorspace_2.1-0 filelock_1.0.3 hdf5r_1.3.10
[64] ROCR_1.0-11 readxl_1.4.3 spatstat.data_3.0-4
[67] magrittr_2.0.3 lmtest_0.9-40 readr_2.1.5
[70] ggtree_3.8.2 viridis_0.6.5 later_1.3.2
[73] lattice_0.21-9 spatstat.geom_3.2-9 future.apply_1.11.1
[76] shadowtext_0.1.3 scattermore_1.2 XML_3.99-0.16.1
[79] RcppAnnoy_0.0.22 pillar_1.9.0 nlme_3.1-163
[82] caTools_1.18.2 compiler_4.3.2 beachmat_2.16.0
[85] RSpectra_0.16-1 stringi_1.8.3 tensor_1.5
[88] GenomicAlignments_1.36.0 plyr_1.8.9 crayon_1.5.2
[91] abind_1.4-5 BiocIO_1.10.0 gridGraphics_0.5-1
[94] googledrive_2.1.1 loupeR_1.1.2 locfit_1.5-9.9
[97] graphlayouts_1.1.1 bit_4.0.5 fastmatch_1.1-4
[100] codetools_0.2-19 BiocSingular_1.16.0 bslib_0.6.1
[103] plotly_4.10.4 mime_0.12 splines_4.3.2
[106] fastDummies_1.7.3 dbplyr_2.3.4 sparseMatrixStats_1.12.2
[109] HDO.db_0.99.1 cellranger_1.1.0 blob_1.2.4
[112] utf8_1.2.4 seqLogo_1.66.0 AnnotationFilter_1.24.0
[115] fs_1.6.3 listenv_0.9.1 DelayedMatrixStats_1.22.6
[118] pkgbuild_1.4.4 ggplotify_0.1.2 tibble_3.2.1
[121] Matrix_1.6-4 callr_3.7.5 statmod_1.5.0
[124] tzdb_0.4.0 tweenr_2.0.3 pkgconfig_2.0.3
[127] tools_4.3.2 cachem_1.0.8 RSQLite_2.3.5
[130] viridisLite_0.4.2 DBI_1.2.2 fastmap_1.1.1
[133] scales_1.3.0 grid_4.3.2 ica_1.0-3
[136] shinydashboard_0.7.2 Rsamtools_2.16.0 sass_0.4.9
[139] patchwork_1.2.0 ggprism_1.0.4 BiocManager_1.30.22
[142] dotCall64_1.1-1 RANN_2.6.1 farver_2.1.1
[145] scatterpie_0.2.1 tidygraph_1.3.1 yaml_2.3.8
[148] rtracklayer_1.60.1 cli_3.6.1 purrr_1.0.2
[151] writexl_1.5.0 leiden_0.4.3.1 lifecycle_1.0.4
[154] uwot_0.1.16 presto_1.0.0 bluster_1.10.0
[157] BSgenome.Hsapiens.UCSC.hg38_1.4.5 BiocParallel_1.34.2 annotate_1.78.0
[160] gtable_0.3.4 rjson_0.2.21 ggridges_0.5.6
[163] progressr_0.14.0 ape_5.7-1 parallel_4.3.2
[166] limma_3.56.2 jsonlite_1.8.8 edgeR_3.42.4
[169] RcppHNSW_0.6.0 TFBSTools_1.38.0 bitops_1.0-7
[172] bit64_4.0.5 Rtsne_0.17 yulab.utils_0.1.4
[175] spatstat.utils_3.1-1 BiocNeighbors_1.18.0 CNEr_1.36.0
[178] jquerylib_0.1.4 metapod_1.8.0 GOSemSim_2.26.1
[181] dqrng_0.3.2 shinyjs_2.1.0 SeuratDisk_0.0.0.9020
[184] R.utils_2.12.3 lazyeval_0.2.2 enrichplot_1.20.3
[187] htmltools_0.5.7 GO.db_3.17.0 rappdirs_0.3.3
[190] ensembldb_2.24.1 glue_1.7.0 TFMPvalue_0.0.9
[193] spam_2.10-0 googlesheets4_1.1.1 XVector_0.40.0
[196] RCurl_1.98-1.14 treeio_1.24.3 BSgenome_1.68.0
[199] gridExtra_2.3 JASPAR2020_0.99.10 igraph_2.0.3
[202] R6_2.5.1 tidyr_1.3.1 RcppRoll_0.3.1
[205] GenomicFeatures_1.52.2 cluster_2.1.4 Rhdf5lib_1.22.1
[208] gargle_1.5.2 aplot_0.2.2 DirichletMultinomial_1.42.0
[211] vipor_0.4.7 DelayedArray_0.26.7 tidyselect_1.2.1
[214] ProtGenerics_1.32.0 ggforce_0.4.2 xml2_1.3.6
[217] future_1.33.1 rsvd_1.0.5 munsell_0.5.0
[220] KernSmooth_2.23-22 fgsea_1.26.0 htmlwidgets_1.6.4
[223] RColorBrewer_1.1-3 biomaRt_2.56.1 rlang_1.1.2
[226] spatstat.sparse_3.0-3 spatstat.explore_3.2-7 remotes_2.5.0
[229] fansi_1.0.6 beeswarm_0.4.0

[ieres-amgen]

I'm Interested in using the same reference object; I believe the annoy index needs to be calculated for it still. I used the below code I tried to piece together based off Azimuth reference building scripts, but those vary a lot and so I would certainly appreciate any insight from Azimuth authors themselves:

# Read in the data, set Idents, and run PCA + UMAP
reference_data <- readRDS("pansci_filtered.rds")
Idents(reference_data) <- "Sub_cell_type_organ"
reference_data <- SCTransform(reference_data, conserve.memory=TRUE, return.only.var.genes = FALSE) # SCTransform to take care of NormalizeData, FindVariableFeatures, and ScaleData
reference_data <- RunPCA(reference_data, npcs = 50) # RunPCA on the SCT residuals
reference_data <- RunUMAP(reference_data, dims=1:50, return.model = TRUE) # Build UMAP model
reference_data <- FindNeighbors(reference_data, dims=1:50) # Find neighbors

# Format object
azimuth.obj <- AzimuthReference(
  object = reference_data,
  refUMAP = "umap",
  refDR = "pca",
  refAssay = "SCT",
  metadata = c("Organ_name", "Main_cell_type", "Sub_cell_type", "Sub_cell_type_organ", "Lineage"),
  dims = 1:50,
  reference.version = "2.0.0"
)

ValidateAzimuthReference(azimuth.obj) #Validate proper reference creation

SaveAnnoyIndex(object=azimuth.obj[["refdr.annoy.neighbors"]], "idx.annoy")
saveRDS(azimuth.obj, "ref.Rds")

[ruggerobelluomo]

@ieres-amgen Thanks for the reply! Do you think that I can subset only heart organ if I am only interested in creating a heart reference? ANd then chaning PCs for UMAP from 50 to maybe 20-30?
reference_data=subset(reference_data,Organ_name=="Heart")
reference_data <- SCTransform(reference_data, conserve.memory=TRUE, return.only.var.genes = FALSE) # SCTransform to take care of NormalizeData, FindVariableFeatures, and ScaleData
...
...

Thanks!

[ieres-amgen]

It seems reasonable to me! But please note I am not affiliated with Seurat/Azimuth in any way, so probably best to take direct advice on this from the package authors.

[ohsumit]

Greetings, if there is any headway on getting a idx.annoy, I would be most interested in it too! Many thanks!