Efficiency metrics are useful in medicinal chemistry to track small molecule progress in lead optimization (LO). Molecular glue degraders are small molecules that mediate targeted protein degradation by chemically inducing proximity between an E3 ligase and a protein target. The potency and depth of protein degradation are important factors in identifying molecular glue drug candidates. We developed degradation efficiency metrics based on both potency and depth of degradation to track lead optimization objectives. We applied these efficiency metrics retrospectively to track optimization of a clinical molecular glue degrader series, resulting in the identification of Golcadomide (CC-99282). This work illustrates that efficiency metrics are beneficial for the identification of molecular glue drug candidates.

Molecular dynamics (MD) simulation is a powerful tool for characterizing ligand-protein conformational dynamics and offers significant advantages over docking and other rigid structure-based computational methods. However, setting up, running, and analyzing MD simulations continues to be a multi-step process making it cumbersome to assess a library of ligands using MD. We present an automated workflow that streamlines setting up, running, and analyzing Desmond MD simulations. The workflow takes a library of pre-docked ligands and a protein structure as input, sets up and runs MD with each protein-ligand complex, and generates simulation fingerprints for each ligand. Simulation fingerprints (SimFP) capture protein-ligand compatibility, including stability of different ligand-pocket interactions and other useful metrics that enable easy rank-ordering of the ligand library for pocket optimization. SimFP from a ligand library can also be used to build machine learning (ML) models that can predict binding assay outcomes and automatically infer important interactions. Unlike relative free-energy methods that are constrained to assess ligands with high chemical similarity, ML models based on SimFPs can accommodate diverse ligand sets. We present a case study on how SimFP helps delineate structure-activity relationship (SAR) trends and explain potency differences across matched-molecular pairs of cyclic peptides targeting the PD-L1 protein.

We present results on the extent to which physics-based simulation (exemplified by FEP+) and focused machine learning (exemplified by QuanSA) are complementary for ligand affinity prediction. For both methods, predictions of activity for LFA-1 inhibitors from a medicinal chemistry lead optimization project were accurate within the applicable domain of each approach. A hybrid model that combined predictions by both approaches by simple averaging performed better than either method, with respect to both ranking and absolute pKi values. Two publicly available FEP+ benchmarks, covering 16 diverse biological targets, were used to test the generality of the synergy. By identifying training data specifically focused on relevant ligands, accurate QuanSA models were derived using ligand activity data known at the time of the original series publications. Results across the 16 benchmark targets demonstrated significant improvements both for ranking and for absolute pKi values using hybrid predictions that combined the FEP+ and QuanSA predicted affinity values. The results argue for a combined approach for affinity prediction that makes use of physics-driven methods as well as those driven by machine learning, each applied carefully on appropriate compounds, with hybrid prediction strategies being employed where possible.

Accurate prediction of intermolecular interaction energies is a fundamental challenge in electronic structure theory due to their subtle character and small magnitudes relative to total molecular energies. Symmetry adapted perturbation theory (SAPT) provides rigorous quantum mechanical means for computing such quantities directly and accurately, but for a computational cost of at least O(N5), where N is the number of atoms. Here, we report machine learned models of SAPT components with a computational cost that scales asymptotically linearly, O(N). We use modified multi-target Behler-Parrinello neural networks and specialized intermolecular symmetry functions to address the idiosyncrasies of the intermolecular problem, achieving 1.2 kcal mol-1 mean absolute errors on a test set of hydrogen bound complexes including structural data extracted from the Cambridge Structural Database and Protein Data Bank, spanning an interaction energy range of 20 kcal mol-1. Additionally, we recover accurate predictions of the physically meaningful SAPT component energies, of which dispersion and induction/polarization were the easiest to predict and electrostatics and exchange-repulsion are the most difficult.

Previous studies have identified a series of imidazo[1,2-a]pyridine (IZP) derivatives as potent allosteric inhibitors of HIV-1 integrase (ALLINIs) and virus infection in cell culture. However, IZPs were also found to be relatively potent activators of the pregnane-X receptor (PXR), raising the specter of induction of CYP-mediated drug disposition pathways. In an attempt to modify PXR activity without affecting anti-HIV-1 activity, rational structure-based design and modeling approaches were used. An X-ray cocrystal structure of (S,S)-1 in the PXR ligand binding domain (LBD) allowed an examination of the potential of rational structural modifications designed to abrogate PXR. The introduction of bulky basic amines at the C-8 position provided macrocyclic IZP derivatives that displayed potent HIV-1 inhibitory activity in cell culture with no detectable PXR transactivation at the highest concentration tested.

We introduce a new method for rapid computation of 3D molecular similarity that combines electrostatic field comparison with comparison of molecular surface-shape and directional hydrogen-bonding preferences (called “eSim”). Rather than employing heuristic “colors” or user-defined molecular feature types to represent conformation-dependent molecular electrostatics, eSim calculates the similarity of the electrostatic fields of two molecules (in addition to shape and hydrogen-bonding). We present detailed virtual screening performance data on the standard 102 target DUD-E set. In its moderately fast screening mode, eSim running on a single computing core is capable of processing over 60 molecules per second. In this mode, eSim performed significantly better than all alternate methods for which full DUD-E data were available (mean ROC area of 0.74, p $$< 10^{-9}$$, by paired t-test, compared with the best performing alternate method). In addition, for 92 targets of the DUD-E set where multiple ligand-bound crystal structures were available, screening performance was assessed using alternate ligands or sets thereof (in their bound poses) as similarity targets. Using the joint alignment of five ligands for each protein target, mean ROC area exceeded 0.82 for the 92 targets. Design-focused application of ligand similarity methods depends on accurate predictions of geometric molecular relationships. We comprehensively assessed pose prediction accuracy by curating nearly 400,000 bound ligand pose pairs across the DUD-E targets. Overall, beginning from agnostic initial poses, we observed an 80% success rate for RMSD $$\le 2.0$$ Å  among the top 20 predicted eSim poses. These examples were split roughly 50/50 into cases with high direct atomic overlap (where a shared scaffold exists between a pair) and low direct atomic overlap (where where a ligand pair has dissimilar scaffolds but largely occupies the same space). Within the high direct atomic overlap subset, the pose prediction success rate was 93%. For the more challenging subset (where dissimilar scaffolds are to be aligned), the success rate was 70%. The eSim approach enables both large-scale screening and rational design of ligands and is rooted in physically meaningful, non-heuristic, molecular comparisons.

Advances in the synthesis and screening of small-molecule libraries have accelerated the discovery of chemical probes for studying biological processes. Still, only a small fraction of the human proteome has chemical ligands. Here, we describe a platform that marries fragment-based ligand discovery with quantitative chemical proteomics to map thousands of reversible small molecule-protein interactions directly in human cells, many of which can be site-specifically determined. We show that fragment hits can be advanced to furnish selective ligands that affect the activity of proteins heretofore lacking chemical probes. We further combine fragment-based chemical proteomics with phenotypic screening to identify small molecules that promote adipocyte differentiation by engaging the poorly characterized membrane protein PGRMC2. Fragmentbased screening in human cells thus provides an extensive proteome-widemap of protein ligandability and facilitates the coordinated discovery of bioactive small molecules and their molecular targets.

Inhibition of signal transduction downstream of the IL-23 receptor represents an intriguing approach to the treatment of autoimmunity. Using a chemogenomics approach marrying kinome-wide inhibitory profiles of a compound library with the cellular activity against an IL-23-stimulated transcriptional response in T lymphocytes, a class of inhibitors was identified that bind to and stabilize the pseudokinase domain of the Janus kinase tyrosine kinase 2 (Tyk2), resulting in blockade of receptor-mediated activation of the adjacent catalytic domain. These Tyk2 pseudokinase domain stabilizers were also shown to inhibit Tyk2-dependent signaling through the Type I interferon receptor but not Tyk2-independent signaling and transcriptional cellular assays, including stimulation through the receptors for IL-2 (JAK1- and JAK3-dependent) and thrombopoietin (JAK2-dependent), demonstrating the high functional selectivity of this approach. A crystal structure of the pseudokinase domain liganded with a representative example showed the compound bound to a site analogous to the ATP-binding site in catalytic kinases with features consistent with high ligand selectivity. The results support a model where the pseudokinase domain regulates activation of the catalytic domain by forming receptor-regulated inhibitory interactions. Tyk2 pseudokinase stabilizers, therefore, represent a novel approach to the design of potent and selective agents for the treatment of autoimmunity.

The design, synthesis and characterization of a phosphonate inhibitor of N-acetylneuraminate-9-phosphate phosphatase (HDHD4) is described. Compound 3, where the substrate C-9 oxygen was replaced with a nonlabile CH2 group, inhibits HDHD4 with a binding affinity (IC50 11 μM) in the range of the native substrate Neu5Ac-9-P (compound 1, Km 47 μM). Combined SAR, modeling and NMR studies are consistent with the phosphonate group in inhibitor 3 forming a stable complex with native Mg2+. In addition to this key interaction, the C-1 carboxylate of the sugar interacts with a cluster of basic residues, K141, R104 and R72. Comparative NMR studies of compounds 3 and 1 with Ca2+ and Mg2+ are indicative of a highly dynamic process in the active site for the HDHD4/Mg2+/3 complex. Possible explanations for this observation are discussed.

We describe an extension to the matched molecular pairs approach that merges pairwise activity differences with three-dimensional contextual information derived from X-ray crystal structures and binding pose predictions. The incorporation of 3D binding poses allows the direct comparison of structural changes to diverse chemotypes in particular binding pockets, facilitating the transfer of SAR from one series to another. Integrating matched pair data with the receptor structure can also highlight activity patterns within the binding site--for example, "hot spot" regions can be visualized where changes in the ligand structure are more likely to impact activity. The method is illustrated using P38α structural and activity data to generate novel hybrid ligands, identify SAR transfer networks, and annotate the receptor binding site.

The purpose of this study was to determine the impact of CYP3A5 expression on inhibitory potency (K-i or IC50 values) of CYP3A inhibitors, using recombinant CYP3A4 and CYP3A5 (rCYP3A4 and rCYP3A5) and CYP3A5 genotyped human liver microsomes (HLMs). IC50 ratios between rCYP3A4 and rCYP3A5 (rCYP3A5/rCYP3A4) of ketoconazole (KTZ) and itraconazole (ITZ) were 8.5 and 8.8 for midazolam (MDZ), 4.7 and 9.1 for testosterone (TST), 1.3 and 2.8 for terfenadine, and 0.6 and 1.7 for vincristine, respectively, suggesting substrate- and inhibitor-dependent selectivity of the two azoles. Due to the difference in the IC50 values for CYP3A4 and CYP3A5, nonconcordant expression of CYP3A4 and CYP3A5 protein can significantly affect the observed magnitude of CYP3A-mediated drug-drug interactions in humans. Indeed, the IC50 values of KTZ and ITZ for CYP3A-catalyzed MDZ and TST metabolism were significantly higher in HLMs with CYP3A5*1/*1 and CYP3A5*1/*3 genotypes than in HLMs with the CYP3A5*3/*3 genotype, showing CYP3A5 expression-dependent IC50 values. Moreover, when IC50 values of KTZ and ITZ for different HLMs were kinetically simulated based on CYP3A5 expression level and enzyme-specific IC50 values, a good correlation between the simulated and the experimentally measured IC50 values was observed. Further simulation analysis revealed that both the Ki ratio (for inhibitors) and V-max/K-m ratio (for substrates) between CYP3A4 and CYP3A5 were major factors for CYP3A5 expression- dependent IC50 values. In conclusion, the present study demonstrates that CYP3A5 genotype and expression level have a significant impact on inhibitory potency for CYP3A-catalyzed drug metabolism, but that the magnitude of its effect is inhibitor-substrate pair specific.

In previous studies, gemfibrozil acyl-β-glucuronide, but not gemfibrozil, was found to be a mechanism-based inhibitor of cytochrome P450 2C8. To better understand whether this inhibition is specific for gemfibrozil acyl-β-glucuronide or whether other glucuronide conjugates are potential substrates for inhibition of this enzyme, we evaluated several pharmaceutical compounds (as their acyl glucuronides) as direct-acting and metabolism-dependent inhibitors of CYP2C8 in human liver microsomes. Of 11 compounds that were evaluated as their acyl glucuronide conjugates, only gemfibrozil acyl-β-glucuronide exhibited mechanism-based inhibition, indicating that CYP2C8 mechanism-based inhibition is very specific to certain glucuronide conjugates. Structural analogs of gemfibrozil were synthesized, and their glucuronide conjugates were prepared to further examine the mechanism of inhibition. When the aromatic methyl groups on the gemfibrozil moiety were substituted with trifluoromethyls, the resulting glucuronide conjugate was a weaker inhibitor of CYP2C8 and mechanism-based inhibition was abolished. However, the glucuronide conjugates of monomethyl gemfibrozil analogs were mechanism-based inhibitors of CYP2C8, although not as potent as gemfibrozil acyl-β-glucuronide itself. The ortho-monomethyl analog was a more potent inhibitor than the meta-monomethyl analog, indicating that CYP2C8 favors the ortho position for oxidation and potential inhibition. Molecular modeling of gemfibrozil acyl-β-glucuronide in the CYP2C8 active site is consistent with the ortho-methyl position being the favored site of covalent attachment to the heme. Moreover, hydrogen bonding to four residues (Ser100, Ser103, Gln214, and Asn217) is implicated.

With the financial and productivity challenges currently facing the pharmaceutical industry, there is constant pressure to justify resources and improve efficiency. With process-driven activities, understanding the contribution of these resources is reasonably straightforward. By contrast, measuring the contribution of knowledge workers is less obvious. Here, we present an impact-oriented approach to assessing the performance of an industrial computer-assisted drug design group. We discuss how these metrics are used to understand and optimize resource allocation in support of drug discovery programs.