In silico methods combined with sensitive ligand-based ¹⁹F NMR screening represent an efficient way for identifying fragments binding to difficult targets with low ligandability and for their subsequent optimization. The talk will describe such synergistic approaches leading to the discovery of new chemical matter for relevant pharmaceutical targets.

In drug discovery, medicinal chemists face the challenge of generating and analyzing large datasets, often exceeding a thousand molecules and numerous physicochemical and biological properties. To address this, we introduced BB-SAR, a new methodology which tackles both data complexity and interpretability, by breaking down molecules into their constituent building blocks (BBs). Establishing a direct correlation between molecules and their constituent BBs enables the association of these BBs with their respective biological and physicochemical properties. This facilitates a more intuitive data analysis and enables the identification of critical trends between molecular features and their associated properties. While individual BBs rarely dictate property behavior, their combinations do. BB-SAR identifies impactful combinations for designing new improved compounds. Additionally, it simplifies traditional medicinal chemistry analysis strategies and enhances the efficiency of drug discovery by providing a more inherent understanding of complex datasets within a concise framework.

In the field of drug discovery, "me-too" approaches play a crucial role in identifying potential candidates for further development. These strategies often focus on discovering new lead molecules with scaffolds distinct from the drug under investigation. Our study evaluates the performance and chemical diversity of molecular analogues identified using Pharos3D, a newly introduced technology.

We tested Pharos3D against a dataset comprising top-selling and recently approved first-in-class drugs, ensuring both contemporary relevance and robustness in our analysis. The search was conducted across three ultra-large combinatorial libraries from commercial suppliers: Enamine, ChemSpace, and Synple Chem, with library sizes ranging from twenty billion to up to thirty trillion accessible molecules.

Our findings demonstrate that Pharos3D effectively identifies well-fitting analogues for the drugs in our dataset. These analogues exhibit significant chemical diversity from the original query molecules and offer promising starting points for new medicinal chemistry programs, particularly in addressing ADMET challenges or intellectual property (IP) considerations.

The path from hit compound to drug candidate introduces many challenges in drug discovery. Navigating through large data sets and identifying optimal compound properties during lead optimization can be a daunting task. MOEsaic is a web-based application for organizing, exploring and analyzing large data sets as well as predicting novel hybrid virtual compounds. MOEsaic provides an intuitive interface which follows medicinal chemistry workflows for interrogating  data sets such as matched molecular pairs (MMP) analysis, R-group profiling, MMP transforms, property models and plots. The latest developments include lipophilic efficiency plots and filtering, heatmaps for virtual compounds' predicted properties using custom QSAR/QSPR models, and a greater range of plots for compound comparisons. In addition, MOEsaic also affords the opportunity to configure, standardize and share data across project groups thus providing a fully integrated analytics infrastructure.

Understanding conformational preferences of small molecules as well as the landscape of energetically accessible conformational states of protein binding site residues are important factors for successful structure-based design. I will highlight examples from two Roche projects where these considerations played a crucial role in the design of improved inhibitors for targets against neurological indications.

In a series of phosphodiesterase 10 (PDE10) inhibitors, analysis of ligand strain by comparing experimentally observed and modeled binding modes with the torsional statistics of entries from the Cambridge Structural Database (CSD) was essential to rationalize the SAR and to guide design efforts. By optimizing both ligand strain and interactions with key residues, potent PDE10 inhibition with different binding modes could be achieved.

In a project targeting monoacylglycerol lipase (MAGL), structural analysis of ensembles of X-ray crystal structures indicated a yet unseen concerted movement of two Leu residues in the MAGL binding site. The structural insights were used to substantially improve MAGL inhibition, but more importantly to reduce the sp2 content of the molecules, and thereby significantly improve their ADME properties. This led to a series of very potent, selective, and in-vivo active MAGL inhibitors.

Oligosaccharides play a pivotal role in human biology, influencing a myriad of processes across physiology and pathology. Among them, the tetrasaccharide sialyl Lewis-X (sLeX) is particularly noteworthy for its critical role in the immune response, especially in mediating the rolling and adhesion of leukocytes along the endothelium through its interaction with the E-selectin lectin. This rolling mechanism is essential for the recruitment of immune cells to sites of inflammation, underscoring the therapeutic potential of targeting this interaction.

LeX, a trisaccharide subunit of sLeX, exhibits a highly preorganized conformation that intriguingly lacks conventional hydrogen bonding. However, due to the poor drug-likeness, converting oligosaccharides like sLeX into viable therapeutics requires a comprehensive understanding of their molecular characteristics. This includes not only identifying key structural features but also finetuning existing stabilization forces, that govern their bioactive conformations.

This presentation will demonstrate how advanced structure-based modeling and computational methods can be utilized to decipher and optimize the key intra- and intermolecular interactions in carbohydrate leads. By refining our ability to manipulate the conformational pre-organization and stabilization forces within oligosaccharides, we can enhance their binding affinity and specificity. Recent successes and ongoing challenges in the development of oligosaccharide-based antagonists targeting E-selectin will be presented, paving the way for novel therapeutic strategies in the treatment of inflammatory and related diseases.

Vladas will share his experiences from key contributions to COVID Moonshot lead molecules to discovering and applying the rules of molecular glue designs. Vladas will share how applying computational techniques helped to guide and inspire molecular designs. Some of the molecules have now entered clinical development. Most recently Vladas has been working at VantAI where he is continuing to explore the machine learning innovations and how they can further assist the chemists in the molecular design process.