Skip to main content

Table 3 The average MSE of predicting test expression values \(\tilde{\mathbf {X}}_{test}\) using different models

From: A Graph Feature Auto-Encoder for the prediction of unobserved node features on biological networks

Method

TF_Net

PPI

Genetics

Features

Graph

Features

Graph

Features

Graph

GCN

7.791 ± 3.550

15.127 ± 2.280

6.208 ± 0.607

11.106 ± 0.52198

5.988 ± 0.696

4.560 ± 0.351

GraphSAGE

0.332 ± 0.160

8.078 ± 2.592

0.265 ± 0.135

2.844 ± 0.349

0.233 ± 0.127

4.466 ± 1.605

GraphConv

0.318 ± 0.154

13.812 ± 4.534

0.308 ± 0.139

3.094 ± 0.431

0.234 ± 0.116

5.226 ± 1.054

FeatGraphConv

0.285 ± 0.135

7.525 ± 2.941

0.244 ± 0.130

5.207 ± 1.476

0.201 ± 0.112

3.414 ± 0.691

LR-embedding

1.583 ± 0.200

2.279 ± 0.403

1.091 ± 0.166

1.453 ± 0.264

1.863 ± 0.332

1.653 ± 0.271

RF-embedding

1.945 ± 0.318

2.150 ± 0.363

1.472 ± 0.267

1.452 ± 0.267

1.883 ± 0.343

1.897 ± 0.351

MLP

0.424 ± 0.170

0.354 ± 0.153

0.332 ± 0.134

LR

0.215 ± 0.126

0.147 ± 0.105

0.108 ± 0.084

RF

0.507 ± 0.143

0.194  ± 0.103

1.882 ± 0.343

  1. (Bold indicates lowest error mean per category of experiments for each group of methods (end-to-end, indirect, or baseline)