Multi-View Factor Analysis

Factor analysis methods for multi-omics data integration.

Performs multi-view factor analysis for multi-omics integration. This is a simplified Bayesian factor model with ARD priors for sparsity. For the full MOFA implementation, use the MOFA2 Bioconductor package.

Usage

multi_view_factor_analysis(
  data_list,
  n_factors = 10,
  max_iter = 1000,
  tol = 1e-05,
  sparsity_prior = TRUE,
  verbose = FALSE
)

Arguments

data_list

List of data matrices (features x samples).

n_factors

Number of latent factors (default: 10).

max_iter

Maximum iterations (default: 1000).

tol

Convergence tolerance (default: 1e-5).

sparsity_prior

Use ARD prior for sparsity (default: TRUE).

verbose

Print progress (default: FALSE).

Value

List with factors (Z), weights (W), variance explained, and convergence info.

Details

This function implements a simplified multi-view factor model: Y_m = W_m * Z^T + E_m, where Y_m is the data for view m, W_m are the loadings, Z are the latent factors shared across views, and E_m is Gaussian noise. ARD (Automatic Relevance Determination) priors are used for automatic factor pruning.

Note

For production use of MOFA, we recommend the MOFA2 Bioconductor package which provides the full implementation with Python backend.

Author

Zaoqu Liu; Email: liuzaoqu@163.com