Journal of Cheminformatics

Table 1 Results of the five models and the consensus model, with each model trained on a distinct training set

From: Prediction of the water solubility by a graph convolutional-based neural network on a highly curated dataset

Model	Split	Validation set			Test set
Model	Split	r²	q²	rmse	r²	q²	rmse
GNN1	1	0.870	0.870	0.735	0.873	0.873	0.728
GNN2	2	0.846	0.842	0.809	0.886	0.882	0.701
GNN3	3	0.839	0.839	0.821	0.881	0.880	0.708
GNN4	4	0.846	0.837	0.846	0.874	0.868	0.741
GNN5	5	0.853	0.841	0.841	0.881	0.871	0.734
Consensus	All				0.901	0.896	0.657
Average		0.850 (± 0.011)	0.845 (± 0.012)	0.807 (± 0.040)	0.878 (± 0.005)	0.875 (± 0.005)	0.722 (± 0.015)

Back to article page

ISSN: 1758-2946

Contact us

Submission enquiries: journalsubmissions@springernature.com