Table 4 Best performing models maximised for AUROC during hyperparameter optimisation

\(S^2\)

0.682 ± 0.011

0.211 ± 0.015

525,296

0.695± 0.005

0.247 ± 0.003

258,098

\(LS^2\)

0.687 ± 0.005

0.248 ± 0.014

637,906

0.704 ± 0.011

0.245 ± 0.019

167,101

\(LS^2 + LL\)

0.703 ± 0.007

0.241 ± 0.005

216,404

0.728 ± 0.011

0.300 ± 0.014

592,528

\(LS^2 + LL + ATI\)

0.700 ± 0.009

0.249 ± 0.018

864,817

0.724 ± 0.007

0.303 ± 0.009

545,474

\(LS^2 + LL + ATD\)

0.695 ± 0.013

0.254 ± 0.014

596,364

0.730 ± 0.008

0.304 ± 0.009

1,504,205

\(LS^3 + LL + ATD\)

0.698 ± 0.007

0.244 ± 0.017

7,015,643

0.734 ± 0.008

0.300 ± 0.009

4,037,181

BoW OH LR

0.699 ± 0.007

0.252 ± 0.006

–

0.690 ± 0.015

0.300 ± 0.015

–

GRU

0.701 ± 0.009

0.241 ± 0.007

795,390

0.731 ± 0.008

0.293 ± 0.017

415,892

GRU + ATD

0.710 ± 0.010

0.275 ± 0.012

668,997

0.733 ± 0.008

0.309 ± 0.012

489,151