Paired RNA and ATAC with WNN • scop

Scope

This small example executes weighted nearest-neighbor analysis on 80 paired cells from the packaged pbmcmultiome_sub object. It needs no downloaded fragments, reference sequence or Python environment. The subset and small feature/dimension settings illustrate the interface; they are not recommended parameters for a complete biological analysis.

For your own paired 10x ARC data, import with scop::ReadMultiomeData("filtered_feature_bc_matrix.h5", genome = "hg38"), then pass the returned object to the same WNN call below. Use the verified genome for your data. Both selected assays must contain the same paired cells; WNN reports incomplete pairing before preprocessing instead of silently intersecting or imputing cells.

WNN combines RNA and ATAC neighborhoods. It does not correct sample batch effects. Supplying batch emits a warning rather than silently implying that correction occurred. Review batch effects separately before interpreting joint clusters. See WNN_integrate() for the method contract.

Run paired preprocessing and WNN

data("pbmcmultiome_sub", package = "scop")
paired <- subset(pbmcmultiome_sub, cells = colnames(pbmcmultiome_sub)[1:80])
#> Loading required namespace: SeuratObject
#> Loading required namespace: Signac
paired <- scop::WNN_integrate(
  srt_merge = paired, assay = c("RNA", "peaks"),
  linear_reduction_dims = 10, linear_reduction_dims_use = 2:8,
  nHVF = 100, neighbor_k = 10, nonlinear_reduction_dims = 2,
  verbose = FALSE
)
#> ℹ [2026-11-12 06:14:32] Skip `log1p()` because `layer = data` is not "counts"
#> ℹ [2026-11-12 06:14:35] Skip `log1p()` because `layer = data` is not "counts"
#> ℹ [2026-11-12 06:14:38] Reorder clusters...
#> ℹ [2026-11-12 06:14:38] Skip `log1p()` because `layer = data` is not "counts"
#> Warning: Key 'RNApcaUMAP2D_' taken, using 'rnaumap2d_' instead
#> Warning: Key 'ATACsvdUMAP2D_' taken, using 'atacumap2d_' instead
SeuratObject::Reductions(paired)
#> [1] "RNApca"        "RNApcaUMAP2D"  "RNAUMAP2D"     "ATACsvd"      
#> [5] "ATACsvdUMAP2D" "ATACUMAP2D"    "ATAClsi"       "WNNUMAP2D"

Explicit assay selection keeps derived chromvar and ACTIVITY assays out of the RNA input. Custom RNA assay names must be supplied explicitly. Gene activity is an accessibility-derived proxy, and chromVAR measures motif-related accessibility deviations; neither is measured RNA.

Inspect reusable outputs

Both the default append mode and append = FALSE retain the RNA PCA, ATAC LSI, complete processed assays (including scaled RNA), WNN graphs, modality weights and clustering results. The RNA assay is active on return. The reduction receipt records the actual cells, peaks and selected dimensions at calculation time.

paired@misc$WNN_reduction_list
#> [1] "RNApca"  "ATAClsi"
paired@misc$WNN_dims_list
#> $rna
#> [1] 2 3 4 5 6 7 8
#> 
#> $atac
#> [1] 2 3 4 5 6 7 8
stopifnot(all(paired@misc$WNN_reduction_list %in% SeuratObject::Reductions(paired)))
head(paired[[]][, c("RNA.weight", "ATAC.weight", "WNNclusters")])
#>                    RNA.weight ATAC.weight WNNclusters
#> AAAGCCCGTGTTGCTT-1  0.5756016   0.4243984           2
#> AAAGCTTGTTAGAGGG-1  0.7922759   0.2077241           1
#> AAAGGCTCAATGCCCG-1  0.5354935   0.4645065           2
#> AAATGGCCAAGGTACG-1  0.5100399   0.4899601           4
#> AACAGGATCGTTAGCG-1  0.5961708   0.4038292           4
#> AACCGGCTCAGGCCTA-1  0.4708434   0.5291566           4
SeuratObject::Misc(paired[["ATAClsi"]], "scop_atac_representation")$dimensions
#> [1] 2 3 4 5 6 7 8
stopifnot(nrow(SeuratObject::LayerData(paired, assay = "RNA", layer = "scale.data")) > 0)

Seurat::DimPlot(paired, reduction = "WNNUMAP2D", group.by = "WNNclusters")

For the common dispatcher, use RunIntegration(srt_merge = paired, batch = NULL, integration_methods = "WNN", assay = c("RNA", "peaks"), ...) with the same analysis parameters. Native counts remain available for a subsequent RunLinkPeaks() analysis, which additionally requires compatible gene coordinates and verified sequence statistics. WNN clusters and peak-gene associations do not by themselves establish causal regulation.

Use normalization_method = "SCT" to normalize RNA with SCTransform; ATAC continues to use TF-IDF and LSI. An explicit HVF vector selects RNA features only. Set nonlinear_reduction = character() for graphs and clusters without an embedding. WNN supports UMAP and force-directed embeddings; unsupported methods are reported before analysis instead of being replaced silently.