Background
Methods
Breast Tissue Microarray
Immunohistochemistry
Validation data analysis
Results
Steps for conducting a Weighted Correlation Network Analysis (WGCNA) of patients
1. Create a patient correlation network from tumor marker expression data
2. Define patient groups (modules) from the patient network
3. Evaluate the utility of WGCNA groups for survival prediction
4. Use classification trees to identify key markers for defining WGCNA groups
Application of WGCNA to a tumor expression array breast cancer data set
Trait | Description |
---|---|
Tumor Size in cm | 78 (5%) |
Median (Range) | 2.3 (0.2 - 9.0) |
25th - 75th Quartile | 1.5 - 3.0 |
Clinical Stage | 82 (0%) |
I | 29 (35%) |
II | 36 (44%) |
III-IVa
| 17 (21%) |
Tumor Grade | 79 (4%) |
I | 22 (27%) |
II | 23 (28%) |
III | 34 (41%) |
Lymph Node+ | 29 (35%) |
ER+ | 61 (74%)b
|
PR+ | 58 (71%) |
HER-2/neu+ | 20 (24%)c
|
Metastasis+ | 32 (39%)d
|
# Deaths | 13 (16%) |
Time in months | 82 (0%) |
Median (Range) | 97 (5, 121) |
25th - 75th Quartile | 70 - 110 |
TMA Marker | 25th | Median | 75th | # (% NA) |
---|---|---|---|---|
14-3-3 σ | 5 | 33 | 63 | 79 (4%) |
Annexin A1 | 0 | 0 | 6 | 82 (0%) |
Beta-catenin | 73 | 90 | 98 | 76 (7%) |
BS106 | 0 | 3 | 30 | 79 (4%) |
BU101 | 67 | 95 | 100 | 78 (5%) |
MED28 expressed in the cytoplasm (cyt) | 42 | 67 | 90 | 77 (6%) |
MED28 expressed in the nucleus (nuc) | 40 | 60 | 80 | 77 (6%) |
Estrogen Receptor α (ER α) | 2 | 18 | 50 | 78 (5%) |
GATA3 | 27 | 79 | 100 | 82 (0%) |
HER-2/neu | 15 | 37 | 77 | 77 (6%) |
Histone H3, acetylated on K4 (H3K4) | 80 | 87 | 95 | 74 (10%) |
Histone H3, acetylated on K18 (H3K18) | 84 | 90 | 97 | 74 (10%) |
Mammaglobin | 17 | 57 | 83 | 80 (2%) |
Na-K ATPase-α | 33 | 58 | 80 | 75 (9%) |
Na-K ATPase-β1 | 40 | 70 | 92 | 74 (10%) |
p53 | 0 | 4 | 23 | 80 (2%) |
Progesterone Receptor (PR) | 0 | 15 | 57 | 78 (5%) |
Prolactin inducible protein | 0 | 0 | 1 | 79 (4%) |
RIN-1 | 85 | 93 | 100 | 79 (4%) |
Smad2 | 83 | 97 | 100 | 79 (4%) |
Smad4 cyt | 63 | 90 | 97 | 78 (5%) |
Smad4 nuc | 50 | 79 | 88 | 78 (5%) |
TGF-β receptor II cyt | 3 | 10 | 35 | 78 (5%) |
TGF-β receptor II nuc | 32 | 53 | 73 | 78 (5%) |
S100A7 | 0 | 0 | 0 | 74 (10%) |
Ying Yang 1 (YY1) | 90 | 95 | 100 | 76 (7%) |
WGCNA* Groups | WGCNA Groups | COX Groups | ||||
---|---|---|---|---|---|---|
Low | Moderate | High | Low | Moderate | High | |
Low
| 51 | 1 | 4 | 40 | 7 | 3 |
Moderate
| 4 | 13 | 0 | 7 | 4 | 6 |
High
| 1 | 0 | 8 | 2 | 1 | 5 |
Totals
|
N
|
Agree
|
Disagree
|
N
|
Agree
|
Disagree
|
82 | 72 (88%) | 10 (12%) | 75 | 49 | 26 (35%) |
Mortality Group | # Deaths/# Patients (% Mortality) | ||
---|---|---|---|
WGCNA | WGCNA* | COX | |
Low
| 3/56 (5.4%) | 3/56 (5.4%) | 2/49 (4.1%) |
Moderate
| 4/14 (22%) | 4/17 (24%) | 4/12 (33%) |
High
| 6/12 (50%) | 6/9 (67%) | 7/14 (50%) |
Log rank p
| 3.9 × 10-4
| 9.1 × 10-6
| 1.6 × 10-4
|
CPH Model Predictors | WGCNA | WGCNA* | COX | |||
---|---|---|---|---|---|---|
HR (CI) | p-value | HR (CI) | p-value | HR (CI) | p-value | |
Moderate Mortality | 3.1 (0.5,19) | 0.220 | 1.5 (0.3,7.3) | 0.635 | 17.5 (1.7,178) | 0.016 |
High Mortality | 5.9 (1.1,31) | 0.037 | 3.8 (0.8,18) | 0.094 | 11.0 (1.2,102) | 0.036 |
Lymph Node Involvement | 0.9 (0.3,3.3) | 0.900 | 1.1 (0.3,4.0) | 0.879 | 1.2 (0.3,5.0) | 0.767 |
Metastases | 5.5 (0.4,72) | 0.200 | 4.4 (0.3,58) | 0.261 | 3.1 (0.2,59) | 0.446 |
Stage | 1.9 (0.5,7.7) | 0.370 | 2.2 (0.5,8.9) | 0.284 | 2.7 (0.4,17) | 0.296 |
Her2+ | 2.9 (0.8,11) | 0.120 | 3.0 (0.8,10) | 0.091 | 1.7 (0.5,6.0) | 0.433 |
# Observations
| 66 | 66 | 60 | |||
Model R
2
(p-value)
| 0.326 (1.2 × 10-4) | 0.306 (1.1 × 10-4) | 0.386 (4.7 × 10-5) |
Comparison of WGCNA* patient groups with a conventional step-wise analysis
Validation analysis of WGCNA* and COX groups in gene expression data sets
# | Data Set | Marker Probeset | Hazard Ratio | P-values | Validates at 0.1 | |||||
---|---|---|---|---|---|---|---|---|---|---|
ATP1B1 | TP53 | TGFBR2 | Mod. | High | Mod. | High | Model | |||
1 | Miller 2005 | 1 | 1 | 1 | 0.88 | 2.35 | 0.693 | 0.053 | 0.173 | yes |
2
|
Miller 2005
|
1
|
1
|
2
|
0.94
|
2.39
|
0.853
|
0.049
|
0.184
|
yes
|
3 | Miller 2005 | 1 | 2 | 1 | 1.10 | 0.93 | 0.777 | 0.884 | 0.944 | no |
4 | Miller 2005 | 1 | 2 | 2 | 1.17 | 0.94 | 0.611 | 0.914 | 0.865 | no |
5 | Miller 2005 | 2 | 1 | 1 | 0.82 | 1.18 | 0.538 | 0.750 | 0.754 | no |
6 | Miller 2005 | 2 | 1 | 2 | 0.88 | 1.20 | 0.676 | 0.727 | 0.840 | no |
7 | Miller 2005 | 2 | 2 | 1 | 1.05 | 0.32 | 0.873 | 0.258 | 0.371 | no |
8 | Miller 2005 | 2 | 2 | 2 | 1.12 | 0.32 | 0.709 | 0.265 | 0.351 | no |
9 | Pawitan 2005 | 1 | 1 | 1 | 0.64 | 3.17 | 0.429 | 0.024 | 0.062 | yes |
10 | Pawitan 2005 | 1 | 1 | 2 | 0.41 | 2.91 | 0.158 | 0.036 | 0.026 | yes |
11
|
Pawitan 2005
|
1
|
2
|
1
|
0.93
|
4.89
|
0.894
|
0.001
|
0.014
|
yes
|
12 | Pawitan 2005 | 1 | 2 | 2 | 0.41 | 4.05 | 0.156 | 0.004 | 0.004 | yes |
13 | Pawitan 2005 | 2 | 1 | 1 | 0.64 | 3.22 | 0.430 | 0.023 | 0.060 | yes |
14 | Pawitan 2005 | 2 | 1 | 2 | 0.41 | 2.95 | 0.159 | 0.034 | 0.025 | yes |
15 | Pawitan 2005 | 2 | 2 | 1 | 0.92 | 4.37 | 0.880 | 0.002 | 0.022 | yes |
16 | Pawitan 2005 | 2 | 2 | 2 | 0.41 | 3.62 | 0.152 | 0.008 | 0.007 | yes |