Search

Recent advancements in deep learning and generative models have significantly expanded the applications of virtual screening for drug-like compounds. Here, we introduce a multitarget transformer model, PCMol , that leverages the latent protein embeddings derived from AlphaFold2 as a means of conditioning a de novo generative model on different targets. Incorporating rich protein representations allows the model to capture their structural relationships, enabling the chemical space interpolation of active compounds and target-side generalization to new proteins based on embedding similarities. In this work, we benchmark against other existing target-conditioned transformer models to illustrate the validity of using AlphaFold protein representations over raw amino acid sequences. We show that low-dimensional projections of these protein embeddings cluster appropriately based on target families and that model performance declines when these representations are intentionally corrupted. We also show that the PCMol model generates diverse, potentially active molecules for a wide array of proteins, including those with sparse ligand bioactivity data. The generated compounds display higher similarity known active ligands of held-out targets and have comparable molecular docking scores while maintaining novelty. Additionally, we demonstrate the important role of data augmentation in bolstering the performance of generative models in low-data regimes. Software package and AlphaFold protein embeddings are freely available at https://github.com/CDDLeiden/PCMol.

Diacylglycerol lipases (DAGL) are responsible for the biosynthesis of the endocannabinoid 2-arachidonoylglycerol. The fluorescent activity-based probes DH379 and HT-01 have been previously shown to label DAGLs and to cross-react with the serine hydrolase ABHD6. Here, we report the synthesis and characterization of two new quenched activity-based probes 1 and 2, the design of which was based on the structures of DH379 and HT-01, respectively. Probe 1 contains a BODIPY-FL and a 2,4-dinitroaniline moiety as a fluorophore-quencher pair, whereas probe 2 employs a Cy5-fluorophore and a cAB40-quencher. The fluorescence of both probes was quenched with relative quantum yields of 0.34 and 0.0081, respectively. The probes showed target inhibition as characterized in activity-based protein profiling assays using human cell- and mouse brain lysates, but were unfortunately not active in living cells, presumably due to limited cell permeability., (© 2018 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim.)