annotation.yaml Reference

This configuration file corresponds to --option=annotation and centralizes the settings for manual annotation, reannotation, cell2location, and RCTD.

Parameter	Default	Description
option	advance_analysis	Stage identifier field stored in the file
results_folder / data_fold / sample_list	results / data / sample.txt	Output directory, data directory, and sample list
run_type / channel	visium / compare_analysis	Platform type and analysis channel
anno_algorithm	manual	Annotation algorithm branch
annotation_list	annotation.txt	Path to the manual mapping file
device	cuda	Device used for model training
max_epochs_reference	250	Number of training epochs for the cell2location reference model
remove_mt	True	Whether to remove mitochondrial genes
N_cells_per_location	30	Prior for the number of cells per location in cell2location
max_epochs_st	30000	Number of training epochs for the cell2location spatial model
shape_type / image_type	False / False	Keywords used to filter spatial layers
image_slice	False	Whether to crop the image region
x1 / x2 / y1 / y2	0	Coordinates of the cropping window
threads	64	Thread setting for RCTD
RCTD_mode	doublet	RCTD running mode
cell_type_col	celltype	Cell type column name in the RCTD reference object
group_by	sample	Grouping column name used for RCTD visualization
max_cores	8	Maximum number of parallel cores for RCTD

Tuning suggestions

1. First decide on anno_algorithm, then tune only the parameters relevant to that branch to avoid mixing settings across methods.

2. For deep-learning-based annotation, first verify the device setting and the training epoch parameters.

3. For the RCTD branch, first confirm that cell_type_col and RCTD_mode are set consistently with the reference data and analysis goal.