James S. Fraser

The SARS CoV 2 accessory protein Orf9b is in a complex monomer-dimer equilibrium that influences its interactions with the host mitochondrial receptor Tom70. This interaction is critical for viral suppression of a Type-1 interferon response during infection. Modulating this equilibrium with a small molecule, either by stabilizing the Orf9b homodimer or blocking its interaction with Tom70, represents a promising strategy for restoring interferon signaling and the antiviral response. To build tool molecules that could test this concept, we performed two screens: a crystallographic fragment screen against the Orf9b homodimer and a high-throughput fluorescence polarization screen for competitors of an Orf9b-derived peptide binding to Tom70. Fragment screening revealed two binding sites with potential to be developed into an inhibitor: one located at the peripheral dimer interface and the other just outside the lipid-binding channel that defines the central dimer interface. Functionalization of the fragments outside of the lipid-binding channel with hydrophobic moieties stabilized the Orf9b homodimer thereby indirectly inhibiting association with Tom70. In parallel, the high throughput screen for competitive inhibitors of the Tom70:Orf9b interaction discovered a separate series of molecules. These molecules display dynamic structure activity relationship (SAR) and could be improved in the future to modulate the interaction between Tom70 and potentially a wide range of substrates. Collectively, these results demonstrate the feasibility of two distinct strategies to manipulate the Orf9b-Tom70 equilibrium, which is critical to the host response to SARS CoV 2 infection.

Recent advances in protein language models, geometric deep learning, and structure prediction frameworks have the potential to accelerate computational drug discovery. However, it remains unclear how well these methods generalize beyond their training sets. Mutation-dependent drug responses present a challenge to these methods because there are often large effects from relatively small perturbations. Here, we systematically benchmark diverse mutational effect and drug-target affinity prediction models using deep mutational scanning datasets spanning 11 inhibitor conditions for the MET kinase domain. Across both mutational effect prediction and drug–target affinity tasks, current AI models produce largely ligand-invariant predictions despite experimentally distinct inhibitor-dependent fitness landscapes. Ligand-aware geometric models, including LigandMPNN and LASErMPNN, partially recover experimentally validated resistance-associated hotspots, highlighting the importance of explicitly modeling protein–ligand interactions and local structural flexibility. In contrast, deep learning affinity prediction models underestimate mutation-induced changes in binding affinity. This deficiency likely reflects mutation-invariant structural priors derived from co-folded structure prediction frameworks and the absence of mutation-conditioned supervision in current affinity datasets. Finally, we further extended our analysis to several kinase-inhibitor deep mutational scanning datasets and confirm the overall trends. Our findings suggest that a primary limitation in improving performance on this task lies in the scarcity of mutation-conditioned protein-ligand training data rather than model architecture alone.

Ligands that bind to the same G protein-coupled receptor (GPCR) can produce widely different levels of signaling. These differences in efficacy are central to pharmacology, yet how a receptor sequence determines signaling level remains incompletely understood. Here we develop a deep mutational pharmacology approach to measure how each residue in the µ-opioid receptor (µOR) shapes responses to 15 opioids, yielding ∼400,000 measurements of variant effects on signaling and expression, including full concentration-response series for three agonists. Mutational effects on potency are strongest near the orthosteric pocket, whereas effects on efficacy extend through the transmembrane core and intracellular face. Ligands share many functional determinants, but sensitivity to mutation differs with efficacy. As efficacy decreases, the residues required for signaling shift from the binding pocket towards the intracellular face; partial agonists therefore depend on residues that are dispensable for full agonists. Mutations can broadly increase receptor responses, allowing antagonists such as naloxone to produce signaling comparable to a full agonist and increasing the efficacy of partial agonist ligands. Combined mutations can also reorder the relative efficacies of ligands, showing that sequence changes reshape pharmacological differences rather than simply increasing or decreasing all responses together. These findings support a model in which ligands of different efficacy activate the receptor through distinct allosteric mechanisms which are determined by residues distributed across the receptor. Mapping these contributions across diverse ligands with deep mutational pharmacology provides a way to understand how one receptor converts molecular recognition into a range of signaling outputs.

To maintain metabolic homeostasis, enzymes must adapt to fluctuating nutrient levels through mechanisms beyond gene expression. Here, we demonstrate that human glutamine synthetase (GS) can reversibly polymerize into filaments aided by a composite binding site formed at the filament interface by the product, glutamine. Time-resolved cryo-electron microscopy (cryo-EM) confirms that glutamine binding stabilizes these filaments, which in turn exhibit reduced catalytic specificity for ammonia at physiological concentrations. This inhibition appears induced by a conformational change that remodulates the active site loop ensemble gating substrate entry. Metadynamics ensemble refinement revealed >10 Å conformational range for the active site loop and that the loop is stabilized by transient contacts. This disorder is significant, as we show that the transient contacts which stabilize this loop in a closed conformation are essential for catalysis both in vitro and in cells. We propose that GS filament formation constitutes a negative-feedback mechanism, directly linking product concentration to the structural and functional remodeling of the enzyme.

Accurate prediction of ligand-bound protein complexes and ranking them by affinity are central problems in drug discovery. While deep-learning co-folding methods can help address these challenges, their evaluation has been hampered by the difficulties in assessing independence from training data and insufficiently large test sets. Here, we test the ability of co-folding methods to predict the structures of 551 ligands, 489 of which are of sufficient quality for evaluation, bound to the SARS-CoV-2 NSP3 macrodomain (Mac1) that were determined after the training cutoff dates. AlphaFold3 (AF3), Boltz-2, and Chai-1 each reproduced >50% of the Mac1 ligand poses to better than 2 Å root-mean-square deviation of experiment. Despite the potential for co-folding to describe protein conformational changes that stabilize ligand binding, we did not find that common conformational rearrangements, including peptide flip and a large loop opening, were recapitulated by the co-folding prediction. For AF3 and Chai-1, ligand pose prediction confidence weakly, but significantly, tracked experimental potency, while DOCK3.7 energies were only weakly correlated. Boltz-2 affinity predictions showed the strongest correlation with measured potency and, after calibration, achieved lower mean absolute error than a baseline predictor. We next assessed whether co-folding scores could rescore docking hit lists to distinguish true ligands from non-binders among hundreds of molecules prospectively experimentally tested against AmpC β-lactamase, the dopamine D4, and the σ ₂ receptors. AF3 ligand pose confidence values did not separate true ligands from high-scoring false positives as effectively as docking scores or Boltz-2 affinity predictions did. Taken together, the modest, but independent correlations of docking score and co-folding confidence or affinity suggest that integrating physics-based and deep-learning approaches may help with hit prioritization and subsequent optimization in structure-based ligand discovery.

Accurate prediction of ligand-bound protein complexes and ranking them by affinity are central problems in drug discovery. While deep learning co-folding methods can help address these challenges, their evaluation has been hampered by the difficulties in assessing independence from training data and insufficiently large test sets. Here we test the ability of co-folding methods to predict the structures of 551 ligands, 489 of which are of sufficient quality for evaluation, bound to the SARS-CoV-2 NSP3 macrodomain (Mac1) that were determined after the training cut-off dates. AlphaFold3 (AF3), Boltz-2, and Chai-1 each reproduced >50% of the Mac1 ligand poses to better than 2 Å RMSD of experiment. Despite the potential for co-folding to describe protein conformational changes that stabilize ligand binding, we did not find that common conformational rearrangements, including peptide flip and a large loop opening, were recapitulated by the co-folding prediction. For AF3 and Chai-1, ligand pose prediction confidence weakly, but significantly, tracked experimental potency, while DOCK3.7 energies were only weakly correlated. Boltz-2 affinity predictions showed the strongest correlation with measured potency and, after calibration, achieved lower mean absolute error than a baseline predictor. We next assessed whether co-folding scores could rescore docking hit-lists to distinguish true ligands from non-binders among hundreds of molecules prospectively experimentally tested against AmpC β-lactamase, the dopamine D4 and the σ2 receptors. AF3 ligand pose confidence values did not separate true ligands from high-scoring false-positives as effectively as docking scores or Boltz-2 affinity predictions did. Taken together, the modest, but independent correlations of docking score and co-folding confidence or affinity suggests that integrating physics-based and deep-learning and approaches may help with hit prioritization and subsequent optimization in structure-based ligand discovery.

Ligand optimization is central to drug discovery, with hundreds of analogues often designed and synthesized between an initial hit and a therapeutic candidate ^1,2 . The efficiency of this process is unclear, partly because there is no random background for optimization to compare against. Such a random background might emerge from systematic random small substitutions across starting ligands, measuring the likelihood of achieving a substantial improvement in affinity or potency, or other property by any single perturbation. Recent literature has suggested that perhaps 10% of analogues with minor modifications improve upon the potency of a parent by tenfold or more ^3,4 , but this number is clouded by reporting bias, intentional improvement and inter-group variability. To begin to establish a background expectation for ligand optimization, here we systematically modified 18 lead molecules across six targets with single-atom changes; 257 compounds were synthesized. Unexpectedly, 11.3% of these random small perturbation analogues improved potency by tenfold or more. Conversely, they typically had worse in vitro pharmacokinetics. Although it was possible to find analogues where the potency increase compensated for inferior exposure and half-life, resulting in more potent compounds in vivo, overall, a frustrated landscape for ligand optimization is revealed. This study begins to establish a background expectation for ligand potency optimization and offers a simple strategy to do so. It also begins to quantify the challenges confronting the field in moving beyond in vitro potency.

Despite redundant cellular pathways to minimize translational errors, errors in protein synthesis are common. Pathways and mechanisms to minimize errors are classified as pre-ribosomal or ribosomal. Pre-ribosomal pathways are primarily concerned with the appropriate charging of tRNAs with their cognate amino acids. By contrast, the ribosomal decoding center is considered ‘blind’ to mischarged tRNAs since these have cognate codon•anti-codon pairing. Here, we identified that in mycobacteria, deletion of the 16S ribosomal RNA methyltransferase gidB led to increased ribosomal discrimination of mischarged tRNAs. Discrimination only occurred in mycobacteria enriched from environments or genetic backgrounds with high rates of mistranslation. GidB deletion was necessary, but not sufficient for reducing mistranslation due to misacylation. Analysis of new cryo-EM structures of the M. smegmatis ribosomes derived from wild-type and gidB -deleted strains point to the interaction between the base methylated by GidB on the 16S RNA and an asparagine on the ribosomal S12 protein that, when mistranslated to aspartate, may be involved in altering translational fidelity. Our data suggest a mechanism by which mycobacterial ribosomes can discriminate mischarged tRNAs and that 16S rRNA differential methylation by GidB may act to prevent catastrophic translational error.

Cytochrome P450 3A4 (CYP3A4) metabolizes roughly half of all marketed drugs, and its inhibition can cause clinically significant drug-drug interactions. The enzyme accommodates chemically diverse ligands, making binding modes and metabolic outcomes difficult to predict. Previous X-ray crystallography efforts have leveraged a truncated construct without the N-terminal segment that tethers CYP3A4 to the membrane. Here we show that the same construct assembles into a symmetric trimer that can be resolved by cryo-EM and determine structures of both unliganded and ligand-bound CYP3A4. Multiple ligands are resolved with density consistent with several mutually exclusive conformations. Protein remodeling to reshape the binding pocket is concentrated in the F/G loop, which is poorly resolved and unmodeled in many X-ray structures. These features likely underlie the poor predictive performance of co-folding methods on this target. The routine use of cryo-EM to resolve CYP3A4 ligand-bound complexes will provide the ground truth data needed to make predictive models of drug metabolism useful in practice.

Systematic mapping of protein-protein interaction (PPI) networks and determining how causal mutations rewire them in autism spectrum disorder (ASD) provide a powerful framework for uncovering disease mechanisms and therapeutic opportunities. Using affinity purification–mass spectrometry, we systematically mapped PPIs for 100 high-confidence ASD genes, uncovering more than 1800 interactions. By assessing the impact of pathogenic missense mutations, leveraging AlphaFold, and validating key findings in human-derived model systems, we identified marked convergence onto shared protein complexes in the wild-type state and convergent PPI rewiring driven by independent mutations. For example, distinct patient-derived variants in FOXP1 disrupt its interactions with FOXP4, leading to changes in cortical neurogenesis and neural activity in brain organoids. Overall, these findings link genetic variation to protein networks and convergent neurodevelopmental dysfunction in ASD.