Table 3.
No. | Gene Name | Screening Phase (in the Talantov Dataset, GEO, GDS1375) | First-Round Validation (in the Raskin Dataset, GEO GSE15605) * | Second-Round Validation (in the Riker Dataset, Oncomine) ** | Novelty (in PubMeds Abstracts) *** | Full in Silico Validation | ||
---|---|---|---|---|---|---|---|---|
63 Patients | 62 Patients | 59 Patients | ||||||
t Test Melanoma vs. Nevi |
Ratio Melan. vs. Nevi | AUC | Validation on AUC Value * | Validation on Ratio Value | Gene Name and Melanoma Words Co-Occurrence | |||
1 | SCNN1A | 1.0 × 10−14 | 0.07 | 0.98 | Yes (0.85) | Yes (−4.94) | 0 | |
2 | ANO1 | 8.6 × 10−18 | 0.231 | 0.98 | Yes (0.87) | |||
3 | GJA1 | 2.1 × 10−9 | 0.086 | 0.98 | Yes (0.88) | |||
4 | GJB5 | 3.4 × 10−13 | 0.1277 | 0.97 | ||||
5 | GJB3 | 2.1 × 10−10 | 0.2453 | 0.96 | Yes (0.86) | Yes (−6.662) | 0 | |
6 | AQP3 | 6.9 × 10−7 | 0.067 | 0.97 | ||||
7 | SCNN1B | 1.4 × 10−14 | 0.313 | 0.96 | Yes (0.87) | |||
8 | VDAC1 | 2.0 × 10−10 | 1.8 | 0.96 | ||||
9 | CLIC1 | 1.9 × 10−12 | 1.738 | 0.96 | ||||
10 | GABRE | 2.3 × 10−13 | 0.203 | 0.95 | Yes (0.88) | Yes (−3.162) | ≥1 | |
11 | KCNK7 | 1.8 × 10−17 | 0.081 | 0.94 | Yes (0.86) | Yes (−2.832) | 0 | |
12 | KCND3 | 2.0 × 10−11 | 0.487 | 0.94 | Yes (0.89) | |||
14 | KCNN4 | 1.5 × 10−6 | 3.79 | 0.93 | ||||
13 | ITPR3 | 1.1 × 10−9 | 2.311 | 0.93 | ||||
15 | KCNK1 | 8.6 × 10−11 | 0.269 | 0.93 | Yes (0.89) | |||
16 | KCNS3 | 7.7 × 10−6 | 0.403 | 0.93 | ||||
17 | TRPM2 | 6.5 × 10−5 | 0.219 | 0.93 | ||||
18 | GRIA1 | 8.7 × 10−11 | 0.1049 | 0.93 | ||||
19 | GJB1 | 3.4 × 10−8 | 2.643 | 0.93 | Yes (0.87) | Yes (3.303) | 0 | |
20 | KCNN2 | 3.6 × 10−10 | 4.337 | 0.91 | Yes (0.91) | Yes (2.284) | 0 |
* validation in GSE15605 was considered effective if AUC > 0.85 (AUC is reported in parenthesis); ** validation in Riker dataset was considered effective if difference of melanoma vs. normal skin is <0.5 or >1.5 (difference is reported in parenthesis, calculated from log2 median-centered intensity according to Oncomine); *** Novelty was assessed onto the five genes validated in the first and second validations steps.