Identification of TB DEGs
A total of 290 DEGs were screened out from GSE34151 dataset, which contained 229 up-regulated and 61 down-regulated genes. The top 10 significantly deregulated genes were presented in Table 1, including IDO1, ISG20, LAD1, EBI3, CCR7, IFI44L, MIR155HG, RGS1, CD86, G0S2, DUOX1, CSF1R, CDH1, FZD2, C1orf162, PALD1, NDRG2, EVI2B, MGAT4A, CHN2.
Table 1. The top 10 up-regulated and down-regulated DEGs
mRNA ID
|
Up/Down
|
LogFC
|
Adj.P-val
|
P-value
|
IDO1
ISG20
LAD1
EBI3
CCR7
IFI44L
MIR155HG
RGS1
CD86
G0S2
|
Up
Up
Up
Up
Up
Up
Up
Up
Up
Up
|
5.62
5.43
5.06
4.99
4.84
4.62
4.55
4.29
4.17
3.97
|
2.77E-177
1.77E-189
4.36E-180
9.91E-155
4.68E-138
1.94E-117
1.72E-148
6.45E-132
8.48E-124
3.15E-158
|
2.34E-181
3.74E-194
2.77E-184
1.89E-158
2.87E-141
4.68E-120
5.10E-152
6.83E-135
1.36E-126
4.00E-162
|
DUOX1
CSF1R
CDH1
FZD2
C1orf162
PALD1
NDRG2
EVI2B
MGAT4A
CHN2
|
Down
Down
Down
Down
Down
Down
Down
Down
Down
Down
|
-2.01
-2.02
-2.03
-2.04
-2.05
-2.06
-2.06
-2.07
-2.08
-2.08
|
2.33E-84
1.81E-61
2.41E-32
6.88E-97
2.19E-50
4.00E-104
5.63E-57
4.15E-78
5.40E-95
6.78E-93
|
3.01E-86
6.08E-63
2.34E-33
5.17E-99
1.14E-51
1.98E-106
2.25E-58
7.12E-80
4.38E-97
6.13E-95
|
Mine of hub genes based on PPI networks
To further mined the genes associated with TB, we constructed PPI network of DEGs and visualized using STRING database and Cytoscape software respectively(Fig.1A). We then screened 7 modules in the PPI network by the MCODE. The most significant module consists of 27 nodes, including IFI44L, MX1, OASL, HERC5, IRF1, RSAD2, IFI27, IFIT2, OAS1, MX2, HERC6, IRF7, IFITM1, OAS2, CXCL10, IFI6, IFITM2, IFI44, IFITM3, GBP1, ISG20, IFIT3, OAS3, IFIT1, ISG15, USP18, DHX58, all of which are increased (Table 2). The Volcano plot and heatmap of hub genes are displayed in Figure.1B and 1C, respectively.
Table 2. Hub genes of cluster 1
SUID
|
MCODE_Cluster
|
MCODE_Score
|
MCODE_Node_Status
|
Gene-Name
|
368
112
113
118
248
120
185
122
124
130
262
134
200
137
153
155
220
349
221
285
226
100
164
101
103
104
492
|
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
Cluster 1
|
21.59384615
21.59384615
21.59384615
21.92028986
22.00000000
21.59384615
21.70666667
21.59384615
21.59384615
21.59384615
21.92028986
21.59384615
21.59384615
21.59384615
21.59384615
21.59384615
21.00000000
21.59384615
22.00000000
21.59384615
21.59384615
21.59384615
21.59384615
21.59384615
21.59384615
21.59384615
22.00000000
|
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Clustered
Seed
|
IFI44L
MX1
OASL
HERC5
IRF1
RSAD2
IFI27
IFIT2
OAS1
MX2
HERC6
IRF7
IFITM1
OAS2
CXCL10
IFI6
IFITM2
IFI44
IFITM3
GBP1
ISG20
IFIT3
OAS3
IFIT1
ISG15
USP18
DHX58
|
Pathway enrichment and biological functions analysis of the DEGs
The 290 DEGs were uploaded to the DAVID 6.8 online tool for pathway enrichment and functional clustering analyses, P-value < 0.05 were represented as significant terms. In regards to biological processes, DEGs were enriched in defense response to virus, type I interferon signaling pathway, and inflammatory response; in regards to Molecular Function, DEGs were enriched in chemokine activity, 2'-5'-oligoadenylate synthetase activity, and cytokine binding; in regards to cellular component, DEGs were enriched in the external side of plasma membrane, cytosol,extracellular space and in regards to KEGG pathway enrichment, DEGs were enriched in Cytokine-cytokine receptor interaction, Toll-like receptor signaling pathway, TNF signaling pathway(Fig.2A). In addition, we further conducted KEGG pathway enrichment analysis on 27 hub genes, we found enrichment of 27 hub genes belonging to the Influenza A, Hepatitis C and RIG-I-like receptor signaling pathway (Fig.2B). Among these, the RIG-I-like receptor signaling pathway was correlated well with the TB infection (Fig.3).
Assess the diagnostic value of hub genes by ROC curve
To screen the hub genes clinical value, we constructed ROC curves of 27 hub genes by using the GraphPad Prism 8.3.0. Among these, the AUC values of each hub gene were >0.500. The top 10 hub genes DHX58,ISG20,IRF1,IRF7,RSAD2,IFI44,IFI44L,USP18,IFITM1,OAS3 with higher diagnostic efficacy were list in Table 3. Among them, we found DHX58 and IRF7 were strongly correlated with the RIG-I-like receptor signaling pathway. Thus, we picked DHX58 and IRF7 out as target genes. We then verified the expression of DHX58 and IRF7 in blood using GSE38456 and found that DHX58 and IRF7 were up-regulated (P<0.01). We constructed ROC curves to compare the screened genes DHX58 and IRF7 in combination with specific antigen A (ESAT-6) and specific antigen B (CFP-10) in T-spot·TB (Fig.4). the results revealed that DHX58+IRF7 (AUC=0.9277, 95%CI:0.8917-0.9637, P<0.01)diagnosis potency was significantly higher than ESAT-6+CFP-10 (AUC=0.7657, 95%CI:0.6984-0.8329, P<0.01). In addition, the sensitivity and specificity values were significantly higher in the DHX58+IRF7 (Sensitivity=88.14, 95%CI:81.07%-92.80%; Specificity=86.67, 95%CI:78.13%-92.21%) compared to the ESAT-6+CFP-10 (Sensitivity=78.21, 95%CI:67.84%-85.92%; Specificity=66.96, 95%CI:57.82%-74.99%). Furthermore, we compared single DHX58 and IRF7 with single ESAT-6 and CFP-10 in T-spot·TB were constructed using ROC curve (Table.4). the result showed the single DHX58 and IRF7 diagnostic capacity were significantly higher than single ESAT-6 and CFP-10.
Table 3. ROC curve analysis of the top 10 hub genes in DC
Items
|
|
|
Performance
|
|
|
|
|
Sensitivity,%
|
95% CI
|
Specificity,%
|
95% CI
|
AUC
|
95% CI
|
ISG20
|
100.00
|
97.13% - 100.00%
|
100.00
|
97.11% - 100.00%
|
1.0000
|
1.0000 - 1.0000
|
DHX58
|
100.00
|
97.13% - 100.00%
|
100.00
|
97.11% - 100.00%
|
1.0000
|
1.0000 - 1.0000
|
IRF1
|
99.23
|
95.77% - 99.96%
|
100.00
|
97.11% - 100.00%
|
0.9999
|
0.9996 - 1.0000
|
IRF7
|
98.46
|
94.56% - 99.73%
|
99.22
|
95.74% - 99.96%
|
0.9989
|
0.9972 - 1.0000
|
IFI44L
|
96.92
|
92.36% - 98.80%
|
99.22
|
95.74% - 99.96%
|
0.9983
|
0.9961 - 1.0000
|
RSAD2
|
100.00
|
97.13% - 100.00%
|
99.22
|
95.74% - 99.96%
|
0.9976
|
0.9929 - 1.0000
|
USP18
|
97.69
|
93.43% - 99.37%
|
98.45
|
94.52% - 99.72%
|
0.9965
|
0.9915 - 1.0000
|
IFITM1
|
96.15
|
91.31% - 98.35%
|
99.22
|
95.74% - 99.96%
|
0.9963
|
0.9925 - 1.0000
|
OAS3
|
90.77
|
84.56% - 94.64%
|
100.00
|
97.11% - 100.00%
|
0.9948
|
0.9902 - 0.9993
|
HERC5
|
97.69
|
93.43% - 99.37%
|
97.67
|
93.39% - 99.37%
|
0.9930
|
0.9857 - 1.0000
|
Table 4. Diagnostic performance of single DHX58, IRF7, ESAT-6, and CFP-10 in patients with TB
Items
|
|
|
Performance
|
|
|
|
Sensitivity, %
|
95% CI
|
Specificity, %
|
95% CI
|
AUC
|
95% CI
|
DHX58
|
84.44
|
71.22%-92.25%
|
86.89
|
76.20%-93.20%
|
0.8893
|
0.8224-0.9561
|
IRF7
|
91.11
|
79.27%-96.49%
|
94.74
|
85.63%-98.57%
|
0.9637
|
0.9310-0.9964
|
ESAT-6
|
79.49
|
64.47%-89.22%
|
71.43
|
58.52%-81.58%
|
0.8017
|
0.7134-0.8901
|
CFP-10
|
76.92
|
61.66%-87.35%
|
62.50
|
49.41%-73.99%
|
0.7305
|
0.6293-0.8318
|
Notes:all the single DHX58, IRF7, ESAT-6, and CFP-10 p-value were <0.01.