Skip to contents

GPU-accelerated SCENIC GRN inference

scenic_grn_sc_gpu.Rd

GPU counterpart of bixverse::scenic_grn_sc(). Runs multi-output tree regression on the WGPU backend and returns a ScenicGrn object. Dispatches on SingleCells (disk-backed .bin counts) or MetaCells (in-memory sparse) from the bixverse package. GBM is not supported on GPU; use the CPU version for learner_type = "grnboost2".

If genes_to_take is NULL, the CPU bixverse::scenic_gene_filter_sc() is used to shortlist targets (cheap min-counts / min-cells scan).

Usage

scenic_grn_sc_gpu(
  object,
  tf_ids,
  scenic_params = bixverse::params_scenic(),
  wave_byte_budget = 4 * 1024^3,
  genes_to_take = NULL,
  cells_to_take = NULL,
  streaming = NULL,
  random_seed = 42L,
  .verbose = TRUE
)

Arguments

object

SingleCells or MetaCells class from bixverse.

tf_ids

Character vector. Gene identifiers of the transcription factors to use as predictors.

scenic_params

Named list. See bixverse::params_scenic().

wave_byte_budget

Numeric. VRAM ceiling for per-wave histogram and cumulative tensors (bytes). Default 4 GiB. Shrink on tight VRAM budgets, raise on 16 GB+ adapters to let the scheduler pick a wider wave.

genes_to_take

Optional character vector. Target genes to include. If NULL, the CPU gene filter runs first.

cells_to_take

Optional character vector. Cell names to include. If NULL, all filtered cells are used.

streaming

Optional boolean. Only used on SingleCells. If TRUE, the streaming GPU driver is used (bounded host memory). If NULL, is auto-picked from cell count via bixverse's internal auto_streaming. Ignored for MetaCells.

random_seed

Integer. For reproducibility.

.verbose

Boolean or integer. Controls verbosity. FALSE -> quiet, TRUE or 1L -> normal, 2L -> detailed.

Value

A ScenicGrn object with the gene x TF importance matrix.

References

Aibar et al., Nat Methods, 2017.