Fine tuning of parameters for particular populations

[adrdufour]

Hello,

bulk entry: 192 samples of bulk RNAseq from 4 conditions, six samples with eight replicates each
scRNAseq: 12 samples from 2 conditions 3-time points with two replicates providing 11 clusters

Codes run perfectly with these commands

bulkseger.scxin <- SCDC_prop(bulk.eset = eset, sc.eset = eset.sc, ct.varname = "seurat_clusters",
                             sample = "SAMP", ct.sub = c("0", "1", "2", "3", "4", "5", "6", "7", "8", "9", "10", "11"))

Unfortunately, I observe that the proportions given to cluster 2 are 0 for every sample. I suspect that those proportions go to clusters 0

Here is my Differential expression test between clusters 0 and 2

gene	p_val	avg_log2FC	pct.1	pct.2
A230046K03Rik	0	0.9424354	0.302	0.080
Atf4	0	0.9636213	0.284	0.068
BC051077	0	1.0934125	0.270	0.051
Btg1	0	1.0234096	0.387	0.126
Btg2	0	2.0308540	0.514	0.085
C3ar1	0	0.9859951	0.344	0.109
Ccnl1	0	1.4244196	0.329	0.047
Cd83	0	1.3736968	0.401	0.121
Cebpb	0	1.6196460	0.465	0.143
Chd4	0	0.8624976	0.495	0.209
Coq4	0	1.1222041	0.251	0.040
Ddx26b	0	1.0279751	0.273	0.061
Ddx3x	0	0.8229021	0.483	0.206
Ddx5	0	0.9960518	0.675	0.320
Dusp1	0	1.8613721	0.384	0.057
Dusp2	0	1.7916930	0.278	0.031
Egr1	0	2.2168216	0.848	0.231
Fos	0	1.8854204	0.568	0.129

And here is my differential between cluster 2 and the other clusters

gene	p_val	avg_log2FC	pct.1	pct.2
Btg2	0	-1.7712867	0.085	0.434
C3ar1	0	-0.8627266	0.109	0.322
Ccnl1	0	-1.0306977	0.047	0.253
Cd83	0	-1.3607788	0.121	0.372
Cebpb	0	-1.6607110	0.143	0.453
Ctsc	0	-1.2052791	0.155	0.435
Dab2	0	-1.8072923	0.171	0.460
Ddx5	0	-0.7751759	0.320	0.616
Dusp1	0	-2.1763548	0.057	0.397
Eef1a1	0	0.4687902	0.987	0.983
Eef2	0	0.8944768	0.649	0.566
Egr1	0	-1.6636821	0.231	0.668
Fos	0	-1.8573273	0.129	0.503

You can see that clusters do not have particular markers distinguishing them from others. The main difference with cluster 0 is the absence of expression of particular markers.

I would like to know if parameter changes to detect this population are possible.

Thanks for your response.
Adrien Dufour.