Skip to main content
. Author manuscript; available in PMC: 2022 Aug 1.
Published in final edited form as: Methods. 2021 Jan 21;192:120–130. doi: 10.1016/j.ymeth.2021.01.004

Table 1.

Model parameters and datasets of each cancer type

BLCA BRCA COAD HNSC KIRC LGG LIHC LUAD LUSC OV SARC STAD UCEC
Number of Total Samples 408 1,080 451 498 532 506 365 511 494 372 259 350 545
Censored Samples- Events 180 152 99 217 173 125 130 185 212 229 98 141 92
Time Range
(Median)		 13–5050
(527)		 1–8605
(866)		 6–4502
(700)		 2–6417
(640)		 2–4537
(1,182)		 1– 6423
(678)		 1–3675
(596)		 4–7248
(664)		 1–5287
(663)		 8– 5481
(1024)		 15– 5723
(947)		 3–7437
(476)		 1–6859
(915)
Clinical Data Age
Weight
Smoking
Family
Clinical
T stage
N stage
M stage		 Age
Clinical
T stage
N stage
M stage		 Age
Weight
Clinical
T stage
N stage
M stage		 Age
Smoking
Clinical
T stage
N stage
M stage		 Age
Clinical
T stage
N stage
M stage		 Age
Family		 Age
Weight
Smoking
Clinical
T stage
N stage
M stage		 Age
Smoking
Clinical
T stage
N stage
M stage		 Age
Smoking
Clinical
T stage
N stage
M stage		 Age
Clinical		 Age Age
Clinical
T stage
N stage
M stage		 Age
Clinical
Weight
Correlation Graph Interactions
(Graph Density)		 1,223,346
(78.95)		 1,183,560
(76.38)		 2,086,718
(134.66)		 1,683,960
(108.67)		 4,016,716
(259.21)		 3,379,034
(218.06)		 4,849,854
(312.97)		 900,192
(58.09)		 916,958
(59.17)		 586,384
(37.84)		 1,378,742
(88.97)		 1,389,548
(89.67)		 1,944,344
(125.47)
Correlation + GeneMania Interactions (Graph Density) 2,135,982
(137.84)		 2,098,548
(135.43)		 2,988,902
(192.88)		 2,592,694
(167.31)		 4,905,018
(316.53)		 4,265,568
(275.27)		 5,710,822
(368.54)		 1,820,950
(117.51)		 1,838,704
(118.66)		 1,510,576
(97.48)		 2,291,796
(147.90)		 2,300,554
(148.46)		 2,839,154
(183.21)
Batch Size 165 216 181 100 213 203 100 205 100 149 104 140 436
L2 Regularization 0.005 0.005 0.01 0.01 0.005 0.005 0.005 0.001 0.01 0.01 0.005 0.01 .005
Hidden Layer Dropout 0.75 0.75 .75 0.9 0.75 0.75 0.95 0.9 0.9 0.9 0.9 0.85 0.9
Number of Nodes in Hidden Layer 512 256 365 365 512 512 365 256 365 256 256 512 256

The table shows all the parameters of the model, a description of the data, and a description of the graphs. Other parameters stay consistent throughout all models. Including GCNN and Cox layer, learning rate, and regularization in the Cox-PH model and Cox-nnet. The time range represents the minimum and maximum sample follow up time or event time in days with the median time in parenthesis. The clinical data used in each model are shown in the table with the determining factor on what was used being what was available. The number of interactions of the correlation graph and the Correlation + GeneMania graph is shown with the density of the network in parenthesis. The following are the tunable parameters of our GCNN.