Fig. 7: Evaluation of transfer learning models with multiple fidelities and a comparison with the multi-fidelity state embedding algorithm. | Nature Communications

Fig. 7: Evaluation of transfer learning models with multiple fidelities and a comparison with the multi-fidelity state embedding algorithm.

From: Transfer learning with graph neural networks for improved molecular property prediction in the multi-fidelity setting

Fig. 7

A Test metrics for QM7b models leveraging three fidelities corresponding to the ZINDO, PBE0, and GW levels of theory (`LoT') and their correlations. B. Evaluation of the transfer learning strategies in a transductive setting and in the context of the established multi-fidelity state embedding (MFSE) method. The multi-fidelity drug discovery datasets are named based on the high-fidelity (DR dose-response) and low-fidelity (SD single dose) datasets. The abbreviations are: AZ AstraZeneca, AID assay identifier, HOMO highest occupied molecular orbital, LUMO lowest unoccupied molecular orbital, MAE mean absolute error. Source data are provided as a Source Data file.

Back to article page