`diffcyt` after `CATALYST` returned NA in `diffcyt-DS-LMM` even when clusters have "sufficient" number of cells across at least half of samples

[denvercal1234GitHub]

Hi there,

In my DS analysis, I checked the number of cells in each cluster (14 clusters) across samples, and they seemed to have enough, but the analysis still returned all NAs in the results.

When cluster 1 was removed (filterSCE(fimmu13803417_live_fcs_data_sce_All_PBMC, k = "meta14", cluster_id %in% c(2:14))), rowData(ds_formula_PBMCCategoryPID_contrastChronicVSAcute_AbundRes$d_counts) %>% as.data.frame() still showed cluster 1 even though now it had count = 0.

Anyone has suggestions for troubleshooting?

#60 mentioned to select a few markers that actually worth testing to compare, but this would not fix the issues I think.

Thank you.

ei_PBMC <- metadata(fimmu13803417_live_fcs_data_sce_All_PBMC)$experiment_info

(ds_formula_PBMCCategoryPID <- diffcyt::createFormula(ei_PBMC, 
    cols_fixed = "Category", 
    cols_random = "PID"))

contrastChronicVSAcute <- diffcyt::createContrast(c(0, 1, 0))

ds_formula_PBMCCategoryPID_contrastChronicVSAcute_AbundRes <- diffcyt(fimmu13803417_live_fcs_data_sce_All_PBMC, 
    formula = ds_formula_PBMCCategoryPID, contrast = contrastChronicVSAcute,
    analysis_type = "DS", method_DS = "diffcyt-DS-LMM",
    clustering_to_use = "meta14", verbose = T)

> rowData(ds_formula_PBMCCategoryPID_contrastChronicVSAcute_AbundRes$res) %>% as.data.frame()

[feanaros]

Hi, do you solve It?