Parallel deconvolution of a bulk expression matrix

For each column \(\boldsymbol{y}_{\cdot i}\) of the bulk matrix \(\boldsymbol{Y}\in\mathcal{M}_{G\times N}\), estimates \(\hat{\boldsymbol{p}}_{\cdot i}\) with every supplied algorithm. When covariance information is provided, DeCovarT methods maximise \(\ell_{\boldsymbol{y}\,|\,\boldsymbol{\zeta}}(\boldsymbol{p})\) under \(\boldsymbol{y}\,|\,(\boldsymbol{\zeta},\boldsymbol{p})\sim \mathcal{N}_{G}( \boldsymbol{\mu}\boldsymbol{p}, \boldsymbol{\Sigma}(\boldsymbol{p}) )\).

Usage

deconvolute_ratios(
  signature_matrix,
  bulk_expression,
  true_ratios = NULL,
  Sigma = NULL,
  deconvolution_functions = NULL,
  scaled = FALSE,
  cores = ifelse(.Platform$OS.type == "unix", getOption("mc.cores",
    parallel::detectCores()), 1)
)

Arguments

signature_matrix

Mean signature \(\boldsymbol{\mu}\in\mathcal{M}_{G\times J}\) (rownames = genes, colnames = cell types). Used as a frequentist plug-in for unobserved latent cell-type profiles \(\boldsymbol{x}_{\cdot j}\).

bulk_expression

Bulk matrix \(\boldsymbol{Y}\in\mathcal{M}_{G\times N}\).

true_ratios

Optional ground-truth proportions for scoring.

Sigma

Optional array \((\boldsymbol{\Sigma}_j)_{j}\in\mathcal{M}_{G\times G\times J}\).

deconvolution_functions

Named list; each element has FUN and optional additional_parameters.

scaled

If TRUE, apply a log2 transform before estimation.

cores

Number of parallel workers.

Value

A tibble of estimated \(\hat{\boldsymbol{p}}\) and metrics, after repair_simplex().

See also

deconvolute_ratios_Marquardt_Levenberg()

Examples

set.seed(1)
genes <- paste0("g", 1:2)
cts <- paste0("ct", 1:2)
mu <- matrix(c(20, 22, 22, 20), nrow = 2, dimnames = list(genes, cts))
Sigma <- array(
  c(1, 0, 0, 1, 1, 0, 0, 1),
  dim = c(2, 2, 2),
  dimnames = list(genes, genes, cts)
)
Y <- simulate_bulk_mixture(mu, Sigma, p = c(0.5, 0.5), n = 2)$Y
deconvolute_ratios(
  signature_matrix = mu,
  bulk_expression = Y,
  true_ratios = c(0.5, 0.5),
  Sigma = Sigma,
  deconvolution_functions = list(
    "nnls" = list(FUN = deconvolute_ratios_nnls)
  ),
  cores = 1
)
#> # A tibble: 2 × 10
#>   model_mse model_rmse model_mae model_coef_determination model_coef_determina…¹
#>       <dbl>      <dbl>     <dbl>                    <dbl>                  <dbl>
#> 1   0.00650     0.0806    0.0806                        0                      0
#> 2   0.0483      0.220     0.220                         0                      0
#> # ℹ abbreviated name: ¹​model_coef_determination_adjusted
#> # ℹ 5 more variables: model_cor <dbl>, ct1 <dbl>, ct2 <dbl>, OMIC_ID <chr>,
#> #   algorithm <chr>