Compute Spatial-Aware Cell Embeddings with Image Information

Source: R/caesar_image.R

This function computes low-dimensional cell embeddings from a gene-by-cell matrix. The method initializes cell embeddings using approximate PCA and refines them through a linear factor model nested a intrinsical conditional autoregressive model.

cellembedding_image_matrix(
  X,
  adjm,
  q = 50,
  reduction.name = "caesar",
  maxIter = 30,
  epsELBO = 1e-06,
  approx_Phi = FALSE,
  verbose = TRUE,
  Phi_diag = TRUE,
  seed = 1
)

Arguments

X

A gene-by-cell matrix (e.g., the `data` slot from a Seurat object) that serves as the input data for dimensional reduction.

adjm

A spatial adjacency matrix representing relationships between cells or spots.

q

An integer specifying the number of dimensions for the reduced embeddings. Default is 50.

reduction.name

A character string specifying the name of the dimensional reduction method. Default is 'caesar'.

maxIter

Maximum number of iterations for the optimization algorithm. Default is 30.

epsELBO

A small number specifying the convergence threshold for the optimization algorithm. Default is 1e-6.

approx_Phi

Logical, indicating whether to use the approximate method for Phi matrix estimation. Default is FALSE.

verbose

Logical, indicating whether to print progress messages. Default is TRUE.

Phi_diag

Logical, indicating whether to constrain the Phi matrix to be diagonal. Default is TRUE.

seed

An integer used to set the random seed for reproducibility. Default is 1.

Value

A matrix containing the computed cell embeddings. The number of rows corresponds to the number of cells, and the number of columns corresponds to the specified number of dimensions (`q`).