Table 4 Performance evaluation on regression datasets from TDC

From: MolGraph-xLSTM as a graph-based dual-level xLSTM framework for enhanced molecular representation and interpretability

 

Caco2

PPBR

LD50

 

RMSE

PCC

RMSE

PCC

RMSE

PCC

FP-GNN

0.408 ± 0.056

0.835 ± 0.027

12.238 ± 1.194

0.614 ± 0.068

0.911 ± 0.040

0.544 ± 0.046

DeeperGCN

0.624 ± 0.034

0.470 ± 0.112

14.634 ± 0.392

0.305 ± 0.040

0.951 ± 0.063

0.472 ± 0.109

DMPNN

0.487 ± 0.103

0.796 ± 0.023

12.497 ± 0.230

0.588 ± 0.031

0.859 ± 0.035

0.608 ± 0.033

HiGNN

0.457 ± 0.064

0.794 ± 0.043

13.247 ± 0.724

0.554 ± 0.056

0.941 ± 0.038

0.523 ± 0.034

TransFoxMol

0.596 ± 0.082

0.719 ± 0.071

13.638 ± 0.349

0.512 ± 0.027

0.922 ± 0.053

0.538 ± 0.066

BiLSTM

0.611 ± 0.051

0.528 ± 0.103

13.930 ± 0.284

0.416 ± 0.041

0.980 ± 0.029

0.446 ± 0.038

AutoML

0.403 ± 0.014

0.820 ± 0.009

13.565 ± 0.139

0.471 ± 0.016

0.841 ± 0.011

0.622 ± 0.012

MolGraph-xLSTM (ours)

0.358 ± 0.015

0.861 ± 0.011

11.772 ± 0.200

0.644 ± 0.019

0.871 ± 0.026

0.600 ± 0.026

  1. Best results are shown in bold.
Back to article page