Geometric Deep Learning for Structure-Based Ligand Design

Abstract

A pervasive challenge in drug design is determining how to expand a ligand�a small molecule that binds to a target biomolecule�in order to improve various properties of the ligand. Adding single chemical groups, known as fragments, is important for lead optimization tasks, and adding multiple fragments is critical for fragmentbased drug design. We have developed a comprehensive framework that uses machine learning and three-dimensional protein−ligand structures to address this challenge. Our method, FRAME, iteratively determines where on a ligand to add fragments, selects fragments to add, and predicts the geometry of the added fragments. On a comprehensive benchmark, FRAME consistently improves predicted affinity and selectivity relative to the initial ligand, while generating molecules with more drug-like chemical properties than docking-based methods currently in widespread use. FRAME learns to accurately describe molecular interactions despite being given no prior information on such interactions. The resulting framework for quality molecular hypothesis generation can be easily incorporated into the workflows of medicinal chemists for diverse tasks, including lead optimization, fragment-based drug discovery, and de novo drug design.

Protein language models are performant in structure-free virtual screening

“…Aggravatingly, recent work also suggests that many SBDD deep learning models merely memorise the ligands, resulting in poor generalisability and outcomes [8]. In the case of models predicting the binding affinity directly, studies have shown that many of these models consider less of the protein-ligand interaction but more so of quantitative structure analysis relationship (QSAR) of the ligand alone in making their predictions [9].…”

Assembly-Induced Membrane Selectivity of Artificial Model Peptides through Entropy–Enthalpy Competition

“…As a data-driven methodology, its outcomes would be impaired by the limited number of available studies or inconsistent data sets resulting from experimental variations. , Additionally, different peptides may possess distinct working mechanisms that are not directly incorporated into existing peptide databases, further impacting the efficacy of machine learning algorithms . In fact, in nearly all machine-learning scenarios, peptides with higher predicted scores may exhibit poorer biological activity in actual tests. − ,, On the other hand, human expertise has been proven helpful in improving peptide performance despite potential biases present within it. ,− Therefore, effectively leveraging the advantages of machine learning to augment human knowledge holds greater promise for advancing peptide-based drug development.…”