Read-depth aware principal component analysis (RdPCA) is a extension of principal component analysis (RdPCA) that extracts biological signals from read-depth to improve dimension reduction in scRNA-seq data analysis.

fit_pca_linear function in model.py returns the RdPCA coordinates.

fit_pca_linear(expr_mat, 
		ndim_lat,
		cc_mat=None,
		size_factor=None,
		batch_mat=None,
		model='poisson',
		device=torch.device('cpu')
		)

• expr_mat: n_sample x n_gene expression matrix
• ndim_lat: number of principal components
• cc_mat: n_sample x 2 cell cycle phase score matrix (e.g. Seurat cell cycle score)
• size_factor: n_sample vector of offset variables (e.g. scran size factors)
• model: currently only poisson is supported
• device: torch device (default: cpu)