Perform MAGIC Imputation — magic

Applies the diffusion operator to impute and denoise data.

Usage

magic_impute(
  data,
  diff_op,
  t = 3,
  t_max = 20,
  threshold = 0.001,
  solver = c("exact", "approximate"),
  verbose = FALSE
)

data: Matrix (cells x genes) to impute.
diff_op: Diffusion operator (Markov matrix).
t: Integer or "auto". Number of diffusion steps.
t_max: Integer. Maximum t for automatic selection. Default is 20.
threshold: Numeric. Convergence threshold for automatic t. Default is 0.001.
solver: Character. "exact" or "approximate". Default is "exact".
verbose: Logical or integer. Verbosity level.

A matrix of imputed values.

MAGIC imputation applies the diffusion operator repeatedly: $$X_{imputed} = P^t \times X$$

When t = "auto", the algorithm stops when the Procrustes disparity between consecutive iterations falls below the threshold.