Skip to main content

Advertisement

Table 2 Comparison of predictive accuracy resulted from different screening methods

From: Identification of hub genes with diagnostic values in pancreatic cancer by bioinformatics analyses and supervised learning methods

Minimum required interaction score Methods Hub genes k Accuracy of kNN algorithm Mean accuracy of random forest algorithm (rerun 100 times)
0.700 Method 1: 724 DGEs-181 candidate genes-genes bearing top 10 degrees in PPI-2 hub genes by survival analysis and cox analysis MMP7, ITGA2 2 78.21% 81.31%
5 84.62%
10 87.18%
23 92.31%
27 93.59%
Method 2: 724 DGEs-181 candidate genes-genes bearing top 10 degrees in PPI ALB, EGF, FN1, ITGA2, COL1A2, SPARC, COL3A1, TIMP1, COL5A1, COL11A1, MMP7 2 79.49% 83.54%
4 70.51%
6 76.92%
9 78.20%
13 80.77%
15 88.46%
18 83.33%
Method 3: 724 DGEs-genes bearing top 10 degrees in PPI-2 hub genes by survival analysis and cox analysis TOP2A, MAD2L1 2 65.38% 69.82%
5 69.23%
8 65.38%
12 66.67%
23 67.95%
Method 4: 724 DGEs-genes bearing top 10 degrees in PPI CCNB1, CCNA2, MAD2L1, TOP2A, UBE2C, CDC20, TTK, MELK, BUB1B, NDC80 2 70.51% 74.81%
5 71.80%
8 76.92%
13 75.64%
23 74.36%
0.400 Method 5: 724 DGEs-181 candidate genes-genes bearing top 10 degrees in PPI-1 hub genes by survival analysis and cox analysis ITGA2 2 74.36% 69.23%
5 80.77%
10 80.77%
14 80.77%
18 82.05%
22 85.90%
Method 6: 724 DGEs-181 candidate genes-genes bearing top 10 degrees in PPI ALB, EGF, ITGA2, FN1, COL1A2, TIMP1, MMP1, COL3A1, PTGS2, CEL 2 82.05% 83.72%
4 71.80%
6 79.49%
10 75.64%
13 74.36%
18 73.08%
Method7:724 DGEs-genes bearing top10 degrees in PPI-2 hub genes by survival analysis and cox analysis TOP2A, MAD2L1 2 65.38% 69.82%
5 69.23%
8 65.38%
12 66.67%
23 67.95%
Method8:724 DGEs-genes bearing top 10 degrees in PPI ALB, GAPDH, EGF, TOP2A, CCNB1, NDC80, CCNA2, CDC20, UBE2C, BUB1B, MAD2L1, TTK, OIP5, KIF11 2 71.79% 73.05%
6 73.08%
11 69.23%
15 70.51%
22 73.08%
  1. Method 1: Identification of DEGs → screening candidate genes with top 25% variance → construction of PPI by candidate genes with minimum required interaction score at 0.700, and further screen candidate genes with top 10 degrees in PPI → Selection of hub genes by survival and cox analyses in TCGA database
  2. Method 2: Identification of DEGs → screening candidate genes with top 25% variance → construction of PPI by candidate genes with minimum required interaction score at 0.700, and further identification of hub genes bearing top 10 degrees in PPI
  3. Method 3: Identification of DEGs → construction of PPI by candidate genes with minimum required interaction score at 0.700, and further screen candidate genes with top 10 degrees in PPI → selection of hub genes by survival and cox analyses in TCGA database
  4. Method 4: Identification of DEGs → construction of PPI by candidate genes with minimum required interaction score at 0.700, and further identification of hub genes bearing top 10 degrees in PPI
  5. Method 5: Identification of DEGs → screening candidate genes with top 25% variance → construction of PPI by candidate genes with minimum required interaction score at 0.400, and further screen candidate genes with top 10 degrees in PPI → selection of hub genes by survival and cox analyses in TCGA database
  6. Method 6: Identification of DEGs → screening candidate genes with top 25% variance → construction of PPI by candidate genes with minimum required interaction score at 0.400, and further identification of hub genes bearing top 10 degrees in PPI
  7. Method 7: Identification of DEGs → construction of PPI by candidate genes with minimum required interaction score at 0.400, and further screen candidate genes with top 10 degrees in PPI → selection of hub genes by survival and cox analyses in TCGA database
  8. Method 8: Identification of DEGs → construction of PPI by candidate genes with minimum required interaction score at 0.400, and further identification of hub genes bearing top 10 degrees in PPI