Skip to contents

Compute LISI scores on a Seurat object

Source: R/LISI.R
RunLISI.Rd

Compute per-cell Local Inverse Simpson's Index (LISI) scores from a dimensional reduction and store them in the meta.data and tools slots of a Seurat object.

Usage

RunLISI(
  srt,
  reductions = NULL,
  reduction = NULL,
  dims = NULL,
  label_colnames = NULL,
  prefix = NULL,
  tool_name = NULL,
  perplexity = 30,
  nn_eps = 0,
  use_rann = TRUE,
  nn_method = c("auto", "rann", "fnn", "exact"),
  tol = 1e-05,
  max_iter = 50,
  overwrite = TRUE,
  verbose = TRUE
)

Arguments

srt

A Seurat object.

reductions

Character vector of dimensional reductions used to compute LISI. If NULL, DefaultReduction() is used.

reduction

Deprecated alias of reductions.

dims

Dimensions to use from the reduction. Default is NULL, which uses all available dimensions.

label_colnames

Character vector of metadata columns used for LISI. If NULL, RunLISI() will try to use srt@misc[["integration_batch"]].

prefix

Prefix used for the stored LISI metadata columns. If NULL, the reduction names are used.

tool_name

Name used to store detailed results in srt@tools. Default is "LISI" when multiple reductions are provided, otherwise paste0(prefix, "_LISI").

perplexity

Effective neighborhood size. Defaults to 30.

nn_eps

Approximation factor passed to RANN::nn2 when RANN is available and use_rann = TRUE. Defaults to 0.

use_rann

Whether to prefer RANN::nn2 over FNN::get.knn when nn_method = "auto" decides not to use the package's built-in exact C++ backend. Defaults to TRUE.

nn_method

Nearest-neighbor backend. Defaults to "auto", which uses a simple heuristic: low-dimensional inputs use the package's exact C++ search, while larger/higher-dimensional inputs fall back to RANN, then FNN, then the built-in exact C++ backend. Set to "hnsw" to use RcppHNSW for a faster, approximate search.

tol

Tolerance used in the binary search for the target perplexity. Defaults to 1e-5.

max_iter

Maximum number of binary-search iterations. Defaults to 50.

overwrite

Whether to overwrite existing metadata columns. Default is TRUE.

verbose

Whether to print the message. Default is TRUE.

Value

A modified Seurat object.

See also

thisutils::compute_lisi, LISIPlot

Examples

data(panc8_sub)
panc8_sub <- integration_scop(
  panc8_sub,
  batch = "tech",
  integration_method = "Harmony5"
)
#> ◌ [2026-04-26 02:22:33] Run integration workflow...
#> Warning: No layers found matching search pattern provided
#> ℹ [2026-04-26 02:22:34] Perform `Seurat::NormalizeData()` on split layers for Seurat v5 integration
#> ℹ [2026-04-26 02:22:36] Perform `Seurat::FindVariableFeatures()` per batch (`HVF_source = 'separate'`)
#> ℹ [2026-04-26 02:22:38] Number of available HVF: 2000
#> Warning: Layer ‘scale.data’ is empty
#> ℹ [2026-04-26 02:22:39] Perform `Seurat::ScaleData()` on split layers for Seurat v5 integration
#> ℹ [2026-04-26 02:22:39] Perform PCA on split layers before `Seurat::IntegrateLayers()`
#> ℹ [2026-04-26 02:22:41] Perform Seurat v5 integration with `HarmonyIntegration()`
#> The `features` argument is ignored by `HarmonyIntegration`.
#> This message is displayed once per session.
#> ℹ [2026-04-26 02:22:42] Estimated dimensions 1:20 for Harmony5
#> ℹ [2026-04-26 02:22:42] Adjust neighbor k from 20 to 20 for small-sample clustering
#> ℹ [2026-04-26 02:22:42] Perform `Seurat::FindClusters()` with "louvain"
#> ℹ [2026-04-26 02:22:43] Reorder clusters...
#> ℹ [2026-04-26 02:22:43] Skip `log1p()` because `layer = data` is not "counts"
#> ℹ [2026-04-26 02:22:43] Perform umap nonlinear dimension reduction using Harmony5 (1:20)
#> ℹ [2026-04-26 02:22:49] Perform umap nonlinear dimension reduction using Harmony5 (1:20)
#> ℹ [2026-04-26 02:22:55] Perform umap nonlinear dimension reduction using pca (1:20)
#> ✔ [2026-04-26 02:23:02] Harmony5 integration completed
names(panc8_sub@reductions)
#> [1] "pca"            "Harmony5"       "Harmony5UMAP2D" "Harmony5UMAP3D"
#> [5] "pcaUMAP2D"     

panc8_sub <- RunLISI(
  panc8_sub,
  reductions = c("pcaUMAP2D", "Harmony5UMAP2D")
)
#> ℹ [2026-04-26 02:23:02] Compute LISI scores from reduction "pcaUMAP2D"
#> ◌ [2026-04-26 02:23:02] Using "exact" nearest-neighbor backend for compute_lisi
#> ℹ [2026-04-26 02:23:02] Compute LISI scores from reduction "Harmony5UMAP2D"
#> ◌ [2026-04-26 02:23:02] Using "exact" nearest-neighbor backend for compute_lisi
#> ✔ [2026-04-26 02:23:02] Stored LISI scores in metadata: "pcaUMAP2D_tech_LISI" and "Harmony5UMAP2D_tech_LISI"
LISIPlot(
  panc8_sub,
  combine = TRUE
)