Skip to main content
Fig. 9 | Journal of Cheminformatics

Fig. 9

From: Generate what you can make: achieving in-house synthesizability with readily available resources in de novo drug design

Fig. 9

Contrasting the generated drug candidate space with the known MGLL ligand space. UMAP visualization of the solved molecular candidate space of 20,055 molecules generated with in-house synthesizability score (“Caspyrus10k”) and target QSAR model as training objectives and known MGLL ligands. Known MGLL ligands are marked as either inactive (“black circle”) or active. Active ligands are differentiated between synthesizable using in-house building blocks (“green circle”) and those that are not (“red circle”). Experimentally tested candidates are denoted with a star. Exemplary scaffolds are highlighted based on the respective cluster most-frequent Murcko scaffolds. UMAP is calculated using Morgan Fingerprints (Radius 3, Size 2048)

Back to article page