Run scTenifoldKnk in-silico knockout analysis

Usage

RunscTenifoldKnk(
  srt,
  gKO,
  assay = NULL,
  layer = "counts",
  features = NULL,
  qc = TRUE,
  qc_mt_threshold = 0.1,
  qc_min_library_size = 1000,
  qc_min_cells = 25,
  nc_lambda = 0,
  nc_nNet = 10,
  nc_nCells = 500,
  nc_nComp = 3,
  nc_scaleScores = TRUE,
  nc_symmetric = FALSE,
  nc_q = 0.9,
  td_K = 3,
  td_maxIter = 1000,
  td_maxError = 1e-05,
  td_nDecimal = 3,
  ma_nDim = 2,
  cores = 1,
  backend = c("r", "cpp"),
  store_networks = TRUE,
  store_manifold = TRUE,
  tool_name = "scTenifoldKnk",
  verbose = TRUE
)

Arguments

srt: A Seurat object.
gKO: Gene symbol or symbols to knock out. All genes must be present after optional feature and QC filtering.
assay: Assay to use. NULL uses the default assay.
layer: Assay layer used as the count matrix.
features: Optional genes to retain before running network construction. If supplied, gKO is always retained when present in the input assay.
qc: Whether to apply scTenifoldKnk-style quality control.
qc_mt_threshold: Maximum mitochondrial read fraction per cell.
qc_min_library_size: Minimum library size per cell.
qc_min_cells: Minimum number of expressing cells required per gene.
nc_lambda, nc_nNet, nc_nCells, nc_nComp, nc_scaleScores, nc_symmetric, nc_q: Network construction parameters forwarded to scTenifoldNet::makeNetworks().
td_K, td_maxIter, td_maxError, td_nDecimal: Tensor decomposition parameters forwarded to scTenifoldNet::tensorDecomposition().
ma_nDim: Manifold-alignment dimension forwarded to scTenifoldNet::manifoldAlignment().
cores: Number of cores used by native network-construction workers and forwarded to downstream linear algebra where applicable.
backend: r calls scTenifoldKnk::scTenifoldKnk() directly and is the default high-consistency path. cpp follows the upstream scTenifoldNet/scTenifoldKnk network construction, tensor decomposition, manifold alignment, and differential-regulation steps while keeping input handling and result storage inside scop.
store_networks: Whether to keep WT/KO tensor networks in srt@tools.
store_manifold: Whether to keep manifold-alignment coordinates in srt@tools.
tool_name: Name of the srt@tools entry.
verbose: Whether to print the message. Default is TRUE.

Value

A Seurat object with scTenifoldKnk results stored in srt@tools[[tool_name]].

Examples

data(pancreas_sub)
gene_use <- "Pdx1"
counts <- GetAssayData5(
  pancreas_sub,
  assay = "RNA",
  layer = "counts"
)
detected <- names(
  sort(Matrix::rowSums(counts > 0),
    decreasing = TRUE
  )
)
features_use <- unique(c(gene_use, head(detected, 300)))

pancreas_sub <- RunscTenifoldKnk(
  pancreas_sub,
  gKO = gene_use,
  features = features_use,
  qc = FALSE,
  nc_nNet = 3,
  nc_nCells = 200,
  td_maxIter = 200,
  store_networks = FALSE,
  store_manifold = TRUE
)
#> ℹ [2026-08-30 05:34:15] Run scTenifoldKnk knockout for "Pdx1" using "r" backend
#> ℹ Building 3 gene regulatory networks (200 cells each)
#> Networks ■■■■■■■■■■■                       33% | ETA:  4s
#> Networks ■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■  100% | ETA:  0s
#> ✔ Network construction complete: 3 networks
#> ℹ [X] Tensor: 301 x 301 x 3 (K=3)
#> ✔ [X] CP decomposition complete (norm explained: 42.6%)
#> ℹ Manifold alignment: 301 shared genes, d=2
#> ✔ Manifold alignment complete: 2 dimensions
#> ✔ [2026-08-30 05:34:26] scTenifoldKnk results stored in `srt@tools[[scTenifoldKnk]]`

dr <- pancreas_sub@tools$scTenifoldKnk$diffRegulation
head(dr)
#>        gene     distance        Z         FC      p.value        p.adj
#> 1      Pdx1 6.777580e-04 3.980535 7621.73506 0.000000e+00 0.000000e+00
#> 33     Gnas 4.318053e-05 1.899597   30.93717 2.665186e-08 4.011106e-06
#> 148    Ssr2 3.983706e-05 1.850710   26.33172 2.875281e-07 2.884865e-05
#> 156  Sec61b 3.779421e-05 1.819095   23.70037 1.125610e-06 8.470212e-05
#> 236    Dad1 3.319124e-05 1.742161   18.27896 1.908025e-05 1.148631e-03
#> 279 Rpl36al 3.094061e-05 1.701186   15.88409 6.734262e-05 3.378355e-03

scTenifoldKnkPlot(pancreas_sub, plot_type = "effect")