Predict cell-type proportions with a trained Scaden ensemble

Uses a fitted Scaden ensemble returned by [scaden()] to predict cell-type proportions for new bulk samples.

Usage

scaden_predict(fit, test_x, device = NULL)

Arguments

fit

A fitted Scaden object returned by [scaden()].

test_x

Numeric matrix with samples in rows and genes in columns.

device

Optional torch device. If `NULL`, the device stored in `fit` is used.

Value

A named list with components:

average_output

Average prediction across the three ensemble models.

prediction_model256

Predictions from the `m256` model.

prediction_model512

Predictions from the `m512` model.

prediction_model1024

Predictions from the `m1024` model.

model256

The fitted `m256` model.

model512

The fitted `m512` model.

model1024

The fitted `m1024` model.

Details

Predictions are generated independently for all three ensemble members and then averaged sample-wise.

Examples

if (FALSE) { # \dontrun{
pred <- scaden_predict(
  fit = fit,
  test_x = proc$test_x
)
} # }