The protein kinase PKMYT1 regulates a key cell cycle checkpoint as part of the cell's DNA-damage repair response, but in cancer, this function can promote tumor cell survival through avoiding mitotic catastrophe. PKMYT1 has been linked to a variety of cancer types, including breast, gastric, and nonsmall cell lung cancers, as well as kidney renal clear cell carcinoma, making it an important therapeutic target. However, potent and selective small-molecule inhibitors of PKMYT1 are scarce, and its specific biological role in tumor proliferation remains understudied. Here, we report the discovery and characterization of a novel PKMYT1 inhibitor, P29, bound to a previously unknown allosteric site. Structural and kinetic data reveal that P29 induces a conformational rearrangement of the P-loop and inhibits PKMYT1 through a mixed ATP competitive and noncompetitive mechanism. A closely related analogue, P32, exhibits selectivity and enhanced potency and engages PKMYT1 in cells. Surprisingly, however, it binds in the ATP binding pocket, demonstrating that subtle chemical modifications can shift binding mode and mechanism of inhibition. Furthermore, computational analysis using structural modeling methods, including AlphaFold2, AlphaFold3, Boltz-2, as well as unbiased MD simulations, indicates that these approaches are limited in their ability to capture this inhibitor-induced cryptic binding site and conformational change. Our study identifies an underexplored allosteric site in PKMYT1 and establishes a new avenue for the rational design of selective kinase inhibitors targeting a cryptic binding site in this emerging drug target. These findings also reveal intrinsic challenges in the computational discovery of noncanonical kinase binding sites and underscore the necessity of integrating computational modeling with experimental testing using structural and functional approaches.

Abstract NBCn2 (SLC4A10), a member of the SLC4 solute carrier (SLC) family, is a sodium-dependent (bi)carbonate transporter that regulates acid extrusion in various brain regions. Mutations in NBCn2 cause severe neurodevelopmental disorders in humans, and knock out studies suggest that its role in regulating neuronal excitability could hold therapeutic potential for seizure disorders such as epilepsy. Despite its physiological importance, NBCn2’s molecular mechanisms remain largely unknown, and there is limited availability of tool compounds to further probe its role in health and disease. Combining cryoEM with computational docking and simulation studies, we herein elucidate NBCn2’s molecular architecture and substrate binding mechanisms on the atomic scale. Via structure-based drug discovery we further identify a compound series that inhibits NBCn2-mediated transport, and characterize its inhibitory mechanisms via cryoEM. Lastly, we showcase the potential of this compound series to template useful probes by demonstrating pharmacological activity both in primary culture as well as brain slices.

ABSTRACT The Na + -dependent citrate transporter NaCT (SLC13A5) is a key regulator of citrate homeostasis and has emerged as a therapeutic target for metabolic and neurological disease, including the SLC13A5 Epilepsy, a rare disease marked by severe sezures and neurodevelopmental delays. Current NaCT inhibitors are substrate-like molecules that competitively bind the substrate binding site. In this study, we identify previously unknown small molecule inhibitors of NaCT by targeting a putative allosteric site located at the dimer interface. We performed a virtual screen of 3.5 million compounds from the ZINC20 database against this site and selected 54 candidates for experimental testing using a cell-based citrate uptake assay. Through initial experiments, we identified three weak inhibitors, and subsequent evaluation of 26 structurally related analogs yielded six compounds with improved potency (IC 50 = 12.78 μM and 15.49 μM). We then performed further analysis of the putative binding site by integrating structural data with deep mutational scanning evidence and comparisons with homolog structures. This analysis highlighted the importance of key residues (e.g., Phe362) in ligand modulation. These findings reveal a promising allosteric pocket and establish a chemically distinct series of NaCT inhibitors, providing a foundation for rational development of pharmacological modulators of NaCT function.

BACKGROUND:Certain prescription drugs used during pregnancy are associated with offspring autism spectrum disorder (ASD). Nonetheless, ASD risk following prenatal exposure to most drugs remains unknown. Furthermore, methodological challenges and ethical concerns hinder the scope for causal inference. METHODS:We used a case-cohort study design of a nationally representative sample from Israel to examine the associations between maternal prescription drug use during pregnancy and offspring ASD. To scrutinize these associations, the analyses were (a) adjusted for indication proxy (level 2 Anatomical Therapeutic Chemical (ATC) codes), (b) repeated using shared pharmacological targets as exposures, and (c) inspected further through target-enrichment analysis. RESULTS:The sample included 1,400 individuals with and 94,713 without an ASD diagnosis. Among all drugs prescribed during pregnancy, five were statistically significantly associated with increased offspring ASD risk after adjustment for indication proxy (e.g., hazard ratio [95% confidence interval] cyproterone = 2.71 [1.17-6.25] and prednisolone = 2.10 [1.27-3.49]), and two with decreased risk (ferrous sulfate = 0.82 [0.68, 0.99] and lynestrenol = 0.43 [0.2, 0.93]). Further analysis revealed four pharmacological targets shared by these drugs, which were themselves associated with ASD (e.g., neuronal acetylcholine receptor α4β4 = 1.45 [1.05-1.99] and serotonin 2b receptor = 1.31 [1.04-1.61]). Enrichment analysis suggested the association between ASD and medications affecting cholinergic and serotonergic signaling. CONCLUSIONS:Increased ASD risk followed prenatal exposure to five prescription drugs, and decreased risk followed exposure to two. Subsequent analyses suggested no confounding by indication in these associations, but further studies are warranted.

Despite rigorous safety evaluations during development, numerous drugs have been withdrawn from the market due to serious toxicities. Here we investigate the features found in drugs with these unanticipated toxicities and apply a machine learning approach to predict if a drug is likely to be withdrawn due to intolerable side effects without the need for human trial data. Our best preforming classifier was an ensemble predictor trained on protein targets, protein structure features, chemical fingerprints, and chemical features that achieved 92% accuracy and 0.845 Matthews Correlation Coefficient with 10-fold holdout test set cross validation. Analysis of features predictive of unanticipated toxicity revealed both known factors such as inhibition of cytochrome P450 as well as yet uninvestigated factors including the inhibition of bile salt export pumps. This predictor and subsequent feature analysis pave the way for the larger role of computational methods in screening potential candidates during drug development.

Current methods for variant effect prediction do not differentiate between pathogenic variants resulting in different disease outcomes and are restricted in application due to a focus on variants with a single molecular consequence. We have developed Variant-to-Phenotype (V2P), a multi-task, multi-output machine learning model to predict variant pathogenicity conditioned on top-level Human Phenotype Ontology disease phenotypes (n = 23) for single nucleotide variants and insertions/deletions throughout the human genome. V2P leverages a unique approach for the modeling of variant effect that incorporates resultant disease phenotypes as output and during training to improve the quality of variant disease phenotype and effect predictions, simultaneously. We describe the architecture, training strategy, and biological features contributing to V2P's output, revealing initial characteristics underlying the relationship between disease genotype and phenotype. Moreover, we demonstrate the benefit of incorporating disease phenotypes for variant effect predictions by comparing V2P with several variant effect predictors across various high-quality evaluation datasets from manually curated databases and functional assays. Finally, we examine how V2P's predictions result in the successful identification of pathogenic variants in real and simulated patient sequencing data, outperforming other tested methods in initial comparisons. V2P offers a complete mapping of human genetic variants to disease-phenotypes, offering a uniquely conditioned set of variant effect characterizations.

G protein-gated inwardly rectifying potassium (GIRK) channels mediate membrane hyperpolarization in response to G protein-coupled receptor activation and are critical for regulating neuronal excitability. The membrane phospholipid phosphatidylinositol 4,5-bisphosphate (PIP2) is essential for regulating a large family of ion channels, and disruptions in PIP2 interactions contribute to some neurological diseases. Structural analyses have identified key residues in PIP2-mediated gating of the GIRK2 channel. Notably, Arginine-92 (R92), a highly conserved basic residue at the membrane interface in GIRK2, interacts with PIP2 as well as cholesteryl hemisuccinate (CHS), a potentiator of GIRK2. Here, we used a combination of electrophysiological assays, fluorescent K+ flux measurements, cryo-electron microscopy, and molecular dynamics simulations, and find that mutations at R92 (Y, F, and Q) not only alter PIP2 sensitivity but can also reveal a novel gating mechanism that is independent of PIP2. These findings indicate that R92 plays a crucial role in modulating GIRK2 channel gating, offering new insights into developing potential therapeutic targets for treating neurological disorders linked to GIRK channel dysfunction. ### Competing Interest Statement The authors have declared no competing interest.

SLC6A14 is a member of the SLC6 family of amino acid transporters and is known for its wide selectivity in transporting various amino acids across the cell membrane. A recent report detailed the Na+ coupling stoichiometry of SLC6A14 as 3 Na+: 1 amino acid substrate, focusing on the transport of the neutral amino acid glycine as the transported substrate. However, it is still unknown how SLC6A14 can also accommodate the transport of amino acids with positively charged side chains. Here, we employed structural modeling and multiple electrophysiological methods to investigate the unique Na+/Cl- coupling mode of SLC6A14. Our results revealed distinct, variable Na+ coupling modes when the transporter was subjected to either cationic or neutral amino acids (+, 0). In addition, our data provide further insight into the kinetic mechanism of SLC6A14, demonstrating that positively charged amino acids are transported with a 1.4- to 4-fold slower turnover rate compared to neutral amino acid substrates. We propose a binding mode in which the positively charged amino acid allows binding of and coupling of transport to only two Na+ ions, with no effect on Cl- coupling. Results from molecular dynamics (MD) simulations are consistent with this proposal. These findings have significant implications for our understanding of the substrate selectivity of the transport process as well as the development of new pharmacological compounds targeting this transporter.

Many drug failures in clinical trials are due to inadequate safety profiles. We developed an in-silico side effect genetic priority score (SE-GPS) that leverages human genetic evidence to inform side effect risk for a given drug target. We construct the SE-GPS in the Open Target dataset using post-marketing side effect data, externally test it in OnSIDES using side effects reported from drug labels and then generate a SE-GPS for 19,422 protein coding genes and 502 phecodes, of which 1.7% had a SE-GPS > 0. To consider drug mechanism, we incorporated the direction of genetic effect into a directional version of the score called the SE-GPS-DOE. We observe that restricting to at least two lines of genetic evidence conferred a 2.3- and 2.5-fold increased risk in side effects in Open Targets and OnSIDES respectively, with increased enrichments in severe drugs. We make all predictions publicly available in a web portal.

Determining the correct direction of effect (DOE), whether to increase or decrease the activity of a drug target, is essential for therapeutic success. We introduce a framework to predict DOE at gene and gene-disease levels using gene and protein embeddings and genetic associations across the allele frequency spectrum, respectively. Specifically, we predict: (1) DOE-specific druggability for 19,450 protein-coding genes with a macro-averaged area under the receiver operating characteristic curve (AUROC) of 0.95; (2) isolated DOE among 2553 druggable genes with a macro-averaged AUROC of 0.85; and (3) gene-disease-specific DOE for 47,822 gene-disease pairs with a macro-averaged AUROC of 0.59, with performance improving with genetic evidence availability. Our predictions outperform existing approaches, are associated with clinical trial success, and identify novel therapeutic opportunities. We uncover genetic and functional differences between activator and inhibitor targets, allowing DOE inference independent of disease context. This framework represents a valuable tool for target selection and drug development.

G protein-gated inwardly rectifying potassium (GIRK) channels mediate membrane hyperpolarization in response to G protein-coupled receptor activation and are critical for regulating neuronal excitability. The membrane phospholipid phosphatidylinositol 4,5-bisphosphate (PIP2) is essential for regulating the large family of inward rectifiers, and disruptions in PIP2 interactions contribute to some neurological diseases. Structural analyses have identified arginine-92 (R92) in GIRK2 as a key amino acid interacting with PIP2 as well as the potentiator cholesteryl hemisuccinate (CHS). Using electrophysiological assays and fluorescent K+ flux measurements, we show that substitutions at R92 (F, Y, or Q) disrupt PIP2 regulation, as well as G protein and alcohol activation. Cryo-EM structures of R92F and R92Q show an unexpected change in the orientation of the slide helix that leads to a "domain swap" between adjacent subunits in the cytoplasmic domain, producing a unique arrangement of the alcohol-binding pocket and G protein-interacting domain. These findings indicate that R92 plays a crucial role in how GIRK2 channel subunits assemble for physiological gating, and likely extend to gating of most inward rectifiers due to the high conservation of arginine in that location.

Studies have shown that drug targets with human genetic support are more likely to succeed in clinical trials. Hence, a tool integrating genetic evidence to prioritize drug target genes is beneficial for drug discovery. We built a genetic priority score (GPS) by integrating eight genetic features with drug indications from the Open Targets and SIDER databases. The top 0.83%, 0.28% and 0.19% of the GPS conferred a 5.3-, 9.9- and 11.0-fold increased effect of having an indication, respectively. In addition, we observed that targets in the top 0.28% of the score were 1.7-, 3.7- and 8.8-fold more likely to advance from phase I to phases II, III and IV, respectively. Complementary to the GPS, we incorporated the direction of genetic effect and drug mechanism into a directional version of the score called the GPS with direction of effect. We applied our method to 19,365 protein-coding genes and 399 drug indications and made all results available through a web portal.

Heterozygous variants in SLC6A1 , encoding the GAT -1 GABA transporter, are associated with seizures, developmental delay, and autism. The majority of affected individuals carry missense variants, many of which are recurrent germline de novo mutations, raising the possibility of gain -of -function or dominant -negative effects. To understand the functional consequences, we performed an in vitro GABA uptake assay for 213 unique variants, including 24 control variants. De novo variants consistently resulted in a decrease in GABA uptake, in keeping with haploinsufficiency underlying all neurodevelopmental phenotypes. Where present, ClinVar pathogenicity reports correlated well with GABA uptake data; the functional data can inform future reports for the remaining 72% of unscored variants. Surface localization was assessed for 86 variants; two-thirds of loss -of -function missense variants prevented GAT -1 from being present on the membrane while GAT -1 was on the surface but with reduced activity for the remaining third. Surprisingly, recurrent de novo missense variants showed moderate loss -of -function effects that reduced GABA uptake with no evidence for dominant -negative or gain -of -function effects. Using linear regression across multiple missense severity scores to extrapolate the functional data to all potential SLC6A1 missense variants, we observe an abundance of GAT -1 residues that are sensitive to substitution. The extent of this missense vulnerability accounts for the clinically observed missense enrichment; overlap with hypermutable CpG sites accounts for the recurrent missense variants. Strategies to increase the expression of the wild -type SLC6A1 allele are likely to be beneficial across neurodevelopmental disorders, though the developmental stage and extent of required rescue remain unknown.

Identifying genetic drivers of chronic diseases is necessary for drug discovery. Here, we develop a machine learning-assisted genetic priority score, which we call ML-GPS, that incorporates genetic associations with predicted disease phenotypes to enhance target discovery. First, we construct gradient boosting models to predict 112 chronic disease phecodes in the UK Biobank and analyze associations of predicted and observed phenotypes with common, rare, and ultra-rare variants to model the allelic series. We integrate these associations with existing evidence using gradient boosting with continuous feature encoding to construct ML-GPS, training it to predict drug indications in Open Targets and externally testing it in SIDER. We then generate ML-GPS predictions for 2,362,636 gene-phecode pairs. We find that the use of predicted phenotypes, which identify substantially more genetic associations than observed phenotypes across the allele frequency spectrum, significantly improves the performance of ML-GPS. ML-GPS increases coverage of drug targets, with the top 1% of all scores providing support for 15,077 gene-phecode pairs that previously had no support. ML-GPS can also identify well-known target-disease relationships, promising targets without indicated drugs, and targets for several drugs in clinical trials, including LRRK2 inhibitors for Parkinson’s disease and olpasiran for cardiovascular disease. Here, the authors introduce ML-GPS, a machine learning framework that prioritizes drug targets for 112 chronic diseases and integrates genetic associations with predicted phenotypes.

Structural features of proteins capture underlying information about protein evolution and function, which enhances the analysis of proteomic and transcriptomic data. Here we develop Structural Analysis of Gene and protein Expression Signatures (SAGES), a method that describes expression data using features calculated from sequence-based prediction methods and 3D structural models. We used SAGES, along with machine learning, to characterize tissues from healthy individuals and those with breast cancer. We analyzed gene expression data from 23 breast cancer patients and genetic mutation data from the COSMIC database as well as 17 breast tumor protein expression profiles. We identified prominent expression of intrinsically disordered regions in breast cancer proteins as well as relationships between drug perturbation signatures and breast cancer disease signatures. Our results suggest that SAGES is generally applicable to describe diverse biological phenomena including disease states and drug effects.

The human L-type amino acid transporter 1 (LAT1; SLC7A5), is an amino acid exchanger protein, primarily found in the blood-brain barrier, placenta, and testis, where it plays a key role in amino acid homeostasis. Cholesterol is an essential lipid that has been highlighted to play a role in regulating the activity of membrane transporters, such as LAT1, yet little is known about the molecular mechanisms driving this phenomenon. Here we perform a comprehensive computational analysis to investigate cholesterol's role in LAT1 structure and function, focusing on four cholesterol-binding sites (CHOL1-4) identified in a recent LAT1-apo inward-open conformation cryo-EM structure. Through a series of independent molecular dynamics (MD) simulations, molecular docking, MM/GBSA free energy calculations, and other analysis tools, we explored the interactions between LAT1 and cholesterol. Our findings suggest that CHOL3 forms the most stable and favorable interactions with LAT1. Principal component analysis (PCA) and center of mass (COM) distance assessments show that CHOL3 binding stabilizes the inward-open state of LAT1 by preserving the spatial arrangement of the hash and bundle domains. Additionally, we propose an alternative cholesterol-binding site for originally assigned CHOL1. Overall, this study improves the understanding of cholesterol's modulatory effect on LAT1 and proposes candidate sites for the discovery of future allosteric ligands with rational design.

ABSTRACT The human L-type amino acid transporter 1 (LAT1; SLC7A5), is an amino acid exchanger protein, primarily found in the blood-brain-barrier, placenta, and testis, where it plays a key role in amino acid homeostasis. Cholesterol is an essential lipid that has been highlighted to play a role in regulating the activity of membrane transporters such as LAT1, yet little is known about the molecular mechanisms driving this phenomenon. Here we perform a comprehensive computational analysis to investigate cholesterol’s role in LAT1 structure and function, focusing on four cholesterol binding sites (CHOL1-4) identified in a recent LAT1-apo inward-open conformation cryo-EM structure. We performed four independent molecular dynamics (MD) simulations of LAT1 bound to each cholesterol molecule, as well as molecular docking, free energy calculation by MM/GBSA, and other analysis tools, to investigate LAT1-cholesterol interactions. Our results indicate that CHOL3 provides the most stable binding interactions with LAT1, and CHOL3 and CHOL1 sites have the largest stabilizing effect on LAT1’s primary functional motifs (hash and bundle) and substrate binding site. Our analysis also uncovers an alternative cholesterol binding site to the originally assigned CHOL1. Our study improves the understanding of cholesterol’s modulatory effect on LAT1 and proposes candidate sites for discovery of future allosteric ligands with rational design.

Interactions between protein kinases and their substrates are critical for the modulation of complex signaling pathways. Currently, there is a large amount of information available about kinases and their substrates in disparate public databases. However, these data are difficult to interpret in the context of cellular systems, which can be facilitated by examining interactions among multiple proteins at once, such as the network of interactions that constitute a signaling pathway. We present KiNet, a user-friendly web portal that integrates and shares information about kinase-substrate interactions from multiple databases of post-translational modifications. KiNet enables the visual exploration of these interactions in systems contexts, such as pathways, domain families, and custom protein set inputs, in an interactive fashion. We expect KiNet to be useful as a knowledge discovery tool for kinase-substrate interactions, and the aggregated KiNet dataset to be useful for protein kinase studies and systems-level analyses. The portal is available at https://kinet.kinametrix.com/ .

Protein kinase function and interactions with drugs are controlled in part by the movement of the DFG and ɑC-Helix motifs that are related to the catalytic activity of the kinase. Small molecule ligands elicit therapeutic effects with distinct selectivity profiles and residence times that often depend on the active or inactive kinase conformation(s) they bind. Modern AI-based structural modeling methods have the potential to expand upon the limited availability of experimentally determined kinase structures in inactive states. Here, we first explored the conformational space of kinases in the PDB and models generated by AlphaFold2 (AF2) and ESMFold, two prominent AI-based protein structure prediction methods. Our investigation of AF2’s ability to explore the conformational diversity of the kinome at various multiple sequence alignment (MSA) depths showed a bias within the predicted structures of kinases in DFG-in conformations, particularly those controlled by the DFG motif, based on their overabundance in the PDB. We demonstrate that predicting kinase structures using AF2 at lower MSA depths explored these alternative conformations more extensively, including identifying previously unobserved conformations for 398 kinases. Ligand enrichment analyses for 23 kinases showed that, on average, docked models distinguished between active molecules and decoys better than random (average AUC (avgAUC) of 64.58), but select models perform well (e.g., avgAUCs for PTK2 and JAK2 were 79.28 and 80.16, respectively). Further analysis explained the ligand enrichment discrepancy between low- and high-performing kinase models as binding site occlusions that would preclude docking. The overall results of our analyses suggested that, although AF2 explored previously uncharted regions of the kinase conformational space and select models exhibited enrichment scores suitable for rational drug discovery, rigorous refinement of AF2 models is likely still necessary for drug discovery campaigns.