Infer gene regulatory networks with GNIPLR

Usage

RunGNIPLR(object, ...)

# S3 method for class 'Seurat'
RunGNIPLR(
  object,
  assay = NULL,
  layer = "counts",
  targets = NULL,
  correlation_threshold = 0.3,
  lasso_degree = 30,
  lasso_alpha = 0.1,
  max_lag = 3,
  backend = c("cpp", "python"),
  max_edges_per_target = Inf,
  output_file = NULL,
  work_dir = tempdir(),
  prefix = "gniplr",
  envname = "scop_env",
  conda = "auto",
  force = FALSE,
  verbose = TRUE,
  ...
)

# S3 method for class 'matrix'
RunGNIPLR(object, ...)

# Default S3 method
RunGNIPLR(
  object,
  targets = NULL,
  genes_in = c("rows", "columns"),
  correlation_threshold = 0.3,
  lasso_degree = 30,
  lasso_alpha = 0.1,
  max_lag = 3,
  backend = c("cpp", "python"),
  max_edges_per_target = Inf,
  output_file = NULL,
  work_dir = tempdir(),
  prefix = "gniplr",
  envname = "scop_env",
  conda = "auto",
  force = FALSE,
  verbose = TRUE,
  ...
)

Arguments

object: A Seurat object or expression matrix.
...: Additional backend-specific arguments.
assay: Assay used when `object` is a Seurat object.
layer: Assay layer used when `object` is a Seurat object.
targets: Optional target genes. If `NULL`, all genes are considered.
correlation_threshold: Relative correlation filter used before lagged regression. A gene pair is tested when its absolute correlation is at least this fraction of the maximum absolute correlation for the regulator.
lasso_degree: Polynomial degree used by the LASSO projection step.
lasso_alpha: LASSO regularization strength used by the projection step.
max_lag: Maximum lag used by Granger-style lagged regression. Values above `3` are capped at `3` to match the GNIPLR reference implementation.
backend: Runtime backend. Supports `"cpp"` and `"python"`.
max_edges_per_target: Maximum incoming regulator edges retained per target. The default `Inf` keeps all positive-importance links.
output_file: Optional path where the adjacency table is written.
work_dir: Working directory used by the Python backend.
prefix: Prefix for temporary files.
envname: Python environment used by the Python backend.
conda: Conda-compatible executable used by the Python backend.
force: Whether to rebuild existing `output_file`.
verbose: Whether to print progress messages.
genes_in: Matrix orientation for matrix inputs. `"rows"` means genes x cells; `"columns"` means cells x genes.

Value

A data frame with columns `TF`, `target`, `importance`, and `pvalue`. The original GNIPLR p-value matrix is stored in `attr(result, "grn_matrix")` when the network is newly inferred.

References

Zhang Y, Chang X, Liu X. Inference of gene regulatory networks using pseudo-time series data. Bioinformatics. 2021;37(16):2423-2431. doi:10.1093/bioinformatics/btab099.

Examples

data(pancreas_sub)
expr <- GetAssayData5(
  pancreas_sub,
  assay = SeuratObject::DefaultAssay(pancreas_sub),
  layer = "counts"
)
expr <- as.matrix(expr[, seq_len(8)])
expr <- expr[
  names(sort(apply(expr, 1, stats::var), decreasing = TRUE))[seq_len(5)],
]

grn <- RunGNIPLR(
  expr,
  genes_in = "rows",
  correlation_threshold = 0,
  lasso_degree = 1,
  max_lag = 1,
  max_edges_per_target = 2
)
#> ℹ [2026-07-28 03:33:41] Running GNIPLR with `backend = cpp` on 5 genes and 8 cells
head(grn)
#>     TF target importance      pvalue
#> 3 Iapp   Nnat  2.2530369 0.005584228
#> 9 Rbp4   Spp1  1.6205718 0.023956766
#> 4  Pyy   Nnat  1.2732768 0.053299510
#> 7 Iapp   Rbp4  1.0860194 0.082031493
#> 8 Nnat   Rbp4  1.0223798 0.094977389
#> 5 Nnat    Pyy  0.9471533 0.112939714
attr(grn, "grn_matrix")[1:3, 1:3]
#>           Iapp       Pyy        Nnat
#> Iapp 0.0000000 0.9380613 0.005584228
#> Pyy  0.8998447 0.0000000 0.053299510
#> Nnat 0.3772449 0.1129397 0.000000000