Intermittent NA Errors in dec_celltype()

[YuxiZhang-0113]

Many thanks for creating spaTalk and particularly its virtual single-cell generation functionality, which has become an essential component of our spatial transcriptomics analysis pipeline. However, we consistently encounter two technical challenges:

Issue 1: Intermittent NAs in dec_celltype() during spot generation

When running:

obj <- dec_celltype(obj, sc_data, sc_celltype, if_doParallel = TRUE, use_n_cores = 20, dec_result = dec_result) # pre-computed deconvolution

We intermittently receive:

Generating single-cell data for each spot
Error in sequence.default(n..1, from = seq.int(s.1, length(df), s.1), :
'from' contains NAs
Calls: dec_celltype ... as.matrix -> as.matrix -> as.matrix.dist -> sequence.default
In addition: Warning message:
In sequence.default(n..1, from = seq.int(s.1, length(df), s.1), :
NAs introduced by coercion to integer range
Execution halted
The error occurs inconsistently (same code may work or fail)

Issue 2: Performance bottlenecks with large datasets

dec_celltype() and dec_cci_all() run extremely slowly for Visium data with only ~2,500 spots

We greatly appreciate your insights on these matters!