Uses a trained OmicsTweezer encoder and predictor to estimate cell-type proportions for real bulk samples.
Usage
omics_predict(
model,
test_x,
celltypes,
samplename = NULL,
batch_size = 128L,
device = c("auto", "cpu", "cuda"),
cuda_index = NULL,
verbose = TRUE
)Arguments
- model
A trained model returned by [omics_train()]. Must contain `encoder` and `predictor`.
- test_x
Numeric matrix with samples in rows and genes in columns.
- celltypes
Character vector of output cell-type names.
- samplename
Optional character vector of sample names for the returned prediction table.
- batch_size
Integer scalar. Prediction batch size.
- device
Character scalar. One of `"auto"`, `"cpu"`, or `"cuda"`. If `"auto"`, CUDA is used when available, otherwise CPU is used.
- cuda_index
Optional integer scalar giving the CUDA device index to use when `device = "cuda"`. CUDA indices are zero-based.
- verbose
Logical scalar. If `TRUE`, prints the selected prediction device.