Estimate useful dimensions from a reduction

Estimate useful dimensions from a reduction

Usage

RunDimsEstimate(
  srt,
  reduction = NULL,
  reduction_method = NULL,
  k = 30L,
  method = c("scree", "intrinsic", "ensemble"),
  min_dims = 5L,
  fallback_max_dims = 50L,
  variance_threshold = 0.8,
  marginal_gain_threshold = 0.5,
  skip_first = FALSE,
  use_stored = TRUE,
  verbose = TRUE
)

Arguments

srt

A Seurat object.

reduction

Name of the dimensional reduction to inspect. Default is NULL, which automatically selects a PCA-like reduction via DefaultReduction() with pattern = "pca".

reduction_method

Optional reduction method name. When set to "nmf" or "glmpca", all available dimensions will be retained.

k

Number of neighbors used by intrinsicDimension::maxLikGlobalDimEst. Default is 30.

method

Dimension-selection method. "scree" uses PCA standard deviations with broken-stick, elbow, cumulative-variance, and marginal-gain criteria. "intrinsic" uses intrinsicDimension::maxLikGlobalDimEst. "ensemble" keeps the larger recommendation from both methods when both are available. Default is "scree".

min_dims

Minimum number of dimensions kept when intrinsic-dimension estimation succeeds. Default is 5.

fallback_max_dims

Maximum number of dimensions kept when no valid estimate is available. Default is 50.

variance_threshold

Cumulative variance threshold used by method = "scree". Default is 0.8.

marginal_gain_threshold

Stop point for marginal variance gain (percentage points) used by method = "scree". Default is 0.5.

skip_first

Whether to drop the first dimension from the returned result. Useful for TFIDF/LSI workflows. Default is FALSE.

use_stored

Whether to use misc$dims_estimate already stored in the reduction when available. Default is TRUE.

verbose

Whether to print the message. Default is TRUE.

Value

An integer vector of dimensions to use.

See also

DimsEstimatePlot

Examples

data(pancreas_sub)
pancreas_sub <- standard_scop(pancreas_sub)
#> ℹ [2026-07-02 09:36:51] Start standard processing workflow...
#> ℹ [2026-07-02 09:36:52] Checking a list of <Seurat>...
#> ! [2026-07-02 09:36:52] Data 1/1 of the `srt_list` is "unknown"
#> ℹ [2026-07-02 09:36:52] Perform `NormalizeData()` with `normalization.method = 'LogNormalize'` on 1/1 of `srt_list`...
#> ℹ [2026-07-02 09:36:52] Perform `FindVariableFeatures()` on 1/1 of `srt_list`...
#> ℹ [2026-07-02 09:36:52] Use the separate HVF from `srt_list`
#> ℹ [2026-07-02 09:36:52] Number of available HVF: 2000
#> ℹ [2026-07-02 09:36:52] Finished check
#> ℹ [2026-07-02 09:36:52] Perform `ScaleData()`
#> ℹ [2026-07-02 09:36:52] Perform pca linear dimension reduction
#> ℹ [2026-07-02 09:36:53] Use stored estimated dimensions 1:23 for Standardpca
#> ℹ [2026-07-02 09:36:53] Perform `Seurat::FindClusters()` with `cluster_algorithm = 'louvain'` and `cluster_resolution = 0.6`
#> ℹ [2026-07-02 09:36:53] Reorder clusters...
#> ℹ [2026-07-02 09:36:53] Skip `log1p()` because `layer = data` is not "counts"
#> ℹ [2026-07-02 09:36:53] Perform umap nonlinear dimension reduction
#> ✔ [2026-07-02 09:37:00] Standard processing workflow completed
RunDimsEstimate(pancreas_sub)
#> ℹ [2026-07-02 09:37:00] Use stored estimated dimensions 1:23 for Standardpca
#>  [1]  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

DimsEstimatePlot(pancreas_sub)