Fig. 3
From: MolPROP: Molecular Property prediction with multimodal language and graph fusion
Latent Embedding Visualization of the MolPROP BACE Classification Model. The learned neural network embeddings of the BACE test set are projected into 2-dimensional space utilizing the UMAP algorithm for A MolPROPGATv2-ChemBERTa-2-77Â M-MLM, B GATv2 (ablated), and C ChemBERTa-2-77Â M-MLM (ablated) models. All panels display the 1st UMAP dimension as the x-axis and the 2nd UMAP dimesion as the y-axis. The 2-dimensional UMAP projection is determined with the 10 nearest neighbors, utilizing the Jaccard distance metric, and a minimum distance of 0.25. The color scheme is displayed in each panel as a binary blue or red circle. The discrete binary values represent the no inhibition (i.e., blue or 0) or inhibition (i.e., red or 1) of human \(\beta\) secretase, BACE