Negative scores with BASE and DCQ methods on deconvolute_mouse function?

[erzakiev]

Hello, so I was wondering if it's normal that both the DCQ and BASE methods produce abundance matrices that contain negative values?

knitr::kable(immunedeconv_results_dcq, digits = 2)

cell_type	25_tpm_ctrl	41_tpm_ctrl	42_tpm_ctrl	44_tpm	45_tpm	46_tpm
B cell	0.02	-0.05	-0.31	0.34	-0.03	0.04
B cell naive	0.01	-0.04	-0.24	0.26	-0.03	0.03
Myeloid dendritic cell	0.03	-0.07	-0.47	0.51	-0.05	0.06
Plasmacytoid dendritic cell	0.00	0.01	0.05	-0.05	0.01	-0.01
Granulocyte-monocyte progenitor	0.00	0.00	-0.01	0.01	0.00	0.00
Macrophage	-0.09	0.21	1.43	-1.54	0.16	-0.18
Stem cell	0.02	-0.05	-0.33	0.35	-0.04	0.04
Monocyte	0.01	-0.02	-0.10	0.11	-0.01	0.01
NK cell	-0.01	0.02	0.11	-0.12	0.01	-0.01
T cell NK	0.00	0.00	0.03	-0.03	0.00	0.00
T cell immature	-0.79	1.90	12.80	-13.75	1.40	-1.56
T cell CD4+ naive	0.00	0.00	0.03	-0.03	0.00	0.00
T cell regulatory (Tregs)	0.00	0.00	0.01	-0.01	0.00	0.00
T cell CD4+ memory	0.00	0.00	0.01	-0.01	0.00	0.00
T cell CD8+ effector memory	0.00	0.01	0.05	-0.05	0.01	-0.01
T cell CD8+ memory	0.00	0.00	0.02	-0.03	0.00	0.00
T cell CD8+ naive	0.00	0.01	0.04	-0.04	0.00	0.00
T cell gamma delta	0.02	-0.05	-0.32	0.34	-0.03	0.04

knitr::kable(immunedeconv_results_base, digits = 2)

cell_type	25_tpm_ctrl	41_tpm_ctrl	42_tpm_ctrl	44_tpm	45_tpm	46_tpm
B cell	0.28	-0.10	0.62	-0.23	-0.33	-0.53
B cell naive	-0.64	1.13	-0.65	-0.49	0.10	0.15
stroma score	-1.35	0.11	-0.66	0.76	1.03	-0.17
Myeloid dendritic cell	1.53	-0.03	-1.28	0.44	-0.89	0.22
Granulocyte-monocyte progenitor	-1.01	-0.60	-0.96	1.54	0.20	0.49
Macrophage	-0.92	0.18	-1.05	1.09	0.14	0.87
Stem cell	0.67	0.91	-0.69	-1.27	-0.27	-0.29
NK cell	-0.35	-1.09	-0.66	0.58	1.11	0.59
T cell immature	-0.58	1.08	-0.56	-0.69	0.13	-0.01
T cell CD4+ naive	1.58	-1.18	0.25	0.31	-0.02	-0.88
T cell CD4+ memory	-0.01	-0.96	-0.67	0.20	1.13	0.46
T cell CD8+ effector memory	0.06	-0.94	-0.94	0.74	0.90	0.42
T cell CD8+ memory	0.46	-1.25	-1.16	0.62	1.06	0.35
T cell CD8+ naive	1.54	-1.16	0.29	0.43	-0.02	-0.88
T cell gamma delta	-0.17	-1.85	-0.05	0.90	0.70	0.47

The input matrix that i use can be found here:
data4immunedeconv.txt

[LorenzoMerotto]

It it normal for these methods to produce negative results since:

• DCQ applies a regression method which is not constrained
• BASE computes a score similar to gene set enrichment for each cell type

So these results should be interpreted just as scores, and the negative values carry information as well.