Table 6 Experimental results on the curated datasets using multi-modal deep learning method.

I

0.7119  ± 0.0178

0.7138  ± 0.0144

0.7239  ± 0.0155

0.7729  ± 0.0167

0.6633  ± 0.0177

0.6857  ± 0.0133

0.7180  ± 0.0171

II

0.7725  ± 0.0063

0.8684  ± 0.0039

0.7178  ± 0.0124

0.7725  ± 0.0063

0.9915  ± 0.0011

0.7689  ± 0.0061

0.0274  ± 0.0065

III

0.9185  ± 0.0049

0.9455  ± 0.0029

0.7610 ± 0.0151

0.9187 ± 0.0049

0.9739 ± 0.0036

0.8976 ± 0.0051

0.1043 ± 0.0191

IV

0.7085 ± 0.0111

0.8002 ± 0.0079

0.6516 ± 0.0196

0.7089 ± 0.0111

0.9185 ± 0.0082

0.6778 ± 0.0106

0.1086 ± 0.0092

patient dropout prediction (regression)

Phase

MAE (↓)

RMSE (↓)

R² (↑)

I

0.4451  ± 0.0030

0.4608  ± 0.0025

0.6284  ± 0.0290

II

0.4203  ± 0.0024

0.4432  ± 0.0020

0.4033  ± 0.0169

III

0.4054  ± 0.0040

0.4285  ± 0.0034

0.4172  ± 0.0154

IV

0.4180  ± 0.0038

0.4385  ± 0.0030

0.2188  ± 0.0318

adverse event prediction (classification)

Phase

PR-AUC (↑)

F1 (↑)

ROC-AUC (↑)

Precision (↑)

Recall (↑)

Accuracy (↑)

Specificity (↑)

I

0.7527  ± 0.0308

0.8038  ± 0.0227

0.8913  ± 0.0213

0.8486  ± 0.0310

0.7641  ± 0.0261

0.8359  ± 0.0191

0.8924  ± 0.0230

II

0.8259  ± 0.0117

0.8695  ± 0.0088

0.7952  ± 0.0128

0.8340  ± 0.0115

0.9081  ± 0.0088

0.7968  ± 0.0120

0.4707  ± 0.0187

III

0.9059  ± 0.0110

0.9297  ± 0.0072

0.8703  ± 0.0166

0.9087  ± 0.0110

0.9519  ± 0.0076

0.8777  ± 0.0114

0.4542  ± 0.0344

mortality event prediction (classification)

Phase

PR-AUC (↑)

F1 (↑)

ROC-AUC (↑)

Precision (↑)

Recall (↑)

Accuracy (↑)

Specificity (↑)

I

0.6531  ± 0.0400

0.7695  ± 0.0300

0.9093  ± 0.0172

0.7587  ± 0.0389

0.7824  ± 0.0401

0.8733  ± 0.0180

0.9074  ± 0.0170

II

0.6628  ± 0.0160

0.6924  ± 0.0140

0.8269  ± 0.0081

0.7876  ± 0.0174

0.6179  ± 0.0163

0.7471  ± 0.0110

0.8576  ± 0.0122

III

0.6685  ± 0.0159

0.6825  ± 0.0188

0.8144  ± 0.0155

0.7880  ± 0.0191

0.6023  ± 0.0224

0.7273  ± 0.0181

0.8459  ± 0.0182

trial approval prediction (classification)

Phase

PR-AUC (↑)

F1 (↑)

ROC-AUC (↑)

Precision (↑)

Recall (↑)

Accuracy (↑)

Specificity (↑)

I

0.5186  ± 0.0206

0.5172  ± 0.0237

0.7649  ± 0.0148

0.6268  ± 0.0268

0.4406  ± 0.0252

0.6443  ± 0.0151

0.7994  ± 0.0190

II

0.4943  ± 0.0136

0.5406  ± 0.0156

0.7358  ± 0.0085

0.6393  ± 0.0158

0.4685  ± 0.0170

0.7086  ± 0.0082

0.8472  ± 0.0088

III

0.6665  ± 0.0117

0.6724  ± 0.0111

0.7298  ± 0.0095

0.7332  ± 0.0126

0.6211  ± 0.0127

0.6630  ± 0.0086

0.7156  ± 0.0148

IV

0.4396  ± 0.0196

0.5797  ± 0.0175

0.6052  ± 0.0151

0.4471  ± 0.0200

0.8258  ± 0.0203

0.5174  ± 0.0146

0.3089  ± 0.0189

drug dose finding (classification)

Phase

PR-AUC (↑)

F1 (↑)

ROC-AUC (↑)

Precision (↑)

Recall (↑)

Accuracy (↑)

Specificity (↑)

II & III

0.5341  ± 0.0124

0.4938  ± 0.0130

0.7586  ± 0.0056

0.5909  ± 0.0183

0.4661  ± 0.0108

0.5807  ± 0.0093

0.8283  ± 0.0034

trial failure reason identification (classification)

Phase

PR-AUC (↑)

F1 (↑)

ROC-AUC (↑)

Precision (↑)

Recall (↑)

Accuracy (↑)

Specificity (↑)

I

0.2545  ± 0.0059

0.1971  ± 0.0061

0.4751  ± 0.0271

0.2249  ± 0.0110

0.2491  ± 0.0038

0.4816  ± 0.0154

0.7494  ± 0.0032

II

0.2949  ± 0.0083

0.1499  ± 0.0027

0.5692  ± 0.0104

0.1071  ± 0.0028

0.2500  ± 1e-6

0.4285  ± 0.0112

0.7499  ± 1e-6

III

0.2785  ± 0.0083

0.1909  ± 0.0108

0.5503  ± 0.0119

0.1943  ± 0.0181

0.2631  ± 0.0070

0.4509  ± 0.0186

0.7529  ± 0.0022

IV

0.2559  ± 0.0066

0.1993  ± 0.0067

0.4766  ± 0.0220

0.2314  ± 0.0103

0.2511  ± 0.0047

0.4815  ± 0.0147

0.7510  ± 0.0041

trial duration prediction (classification)

Phase

MAE (↓)

RMSE (↓)

R² (↑)

I

0.8334 ± 0.0133

1.2611 ± 0.0261

0.6514  ± 0.0085

II

1.2980 ± 0.0202

1.1756 ± 0.0316

0.4125  ± 0.0081

III

1.4411 ± 0.0226

1.8356 ± 0.0302

0.3148  ± 0.0085

eligibility criteria design (generation)

Phase

cosine sim. (↑)

informative (↑)

redundancy (↓)

All

0.6988

0.6518

0.1181