Search

Abstract As the most abundant organic substances in nature, carbohydrates are essential for life. Understanding how carbohydrates regulate proteins in the physiological and pathological processes presents opportunities to address crucial biological problems and develop new therapeutics. However, the diversity and complexity of carbohydrates pose a challenge in experimentally identifying the sites where carbohydrates bind to and act on proteins. Here, we introduce a deep learning model, DeepGlycanSite, capable of accurately predicting carbohydrate-binding sites on a given protein structure. Incorporating geometric and evolutionary features of proteins into a deep equivariant graph neural network with the transformer architecture, DeepGlycanSite remarkably outperforms previous state-of-the-art methods and effectively predicts binding sites for diverse carbohydrates. Integrating with a mutagenesis study, DeepGlycanSite reveals the guanosine-5’-diphosphate-sugar-recognition site of an important G-protein coupled receptor. These findings demonstrate DeepGlycanSite is invaluable for carbohydrate-binding site prediction and could provide insights into molecular mechanisms underlying carbohydrate-regulation of therapeutically important proteins.

Abstract The persistent pandemic of coronavirus disease 2019 (COVID-19) caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and its variants accentuates the great demand for developing effective therapeutic agents. Here, we report the development of an orally bioavailable SARS-CoV-2 3C-like protease (3CLpro) inhibitor, namely simnotrelvir, and its preclinical evaluation, which lay the foundation for clinical trials studies as well as the conditional approval of simnotrelvir in combination with ritonavir for the treatment of COVID-19. The structure-based optimization of boceprevir, an approved HCV protease inhibitor, leads to identification of simnotrelvir that covalently inhibits SARS-CoV-2 3CLpro with an enthalpy-driven thermodynamic binding signature. Multiple enzymatic assays reveal that simnotrelvir is a potent pan-CoV 3CLpro inhibitor but has high selectivity. It effectively blocks replications of SARS-CoV-2 variants in cell-based assays and exhibits good pharmacokinetic and safety profiles in male and female rats and monkeys, leading to robust oral efficacy in a male mouse model of SARS-CoV-2 Delta infection in which it not only significantly reduces lung viral loads but also eliminates the virus from brains. The discovery of simnotrelvir thereby highlights the utility of structure-based development of marked protease inhibitors for providing a small molecule therapeutic effectively combatting human coronaviruses.

Abstract During the ongoing pandemic, providing treatment consisting of effective, low-cost oral antiviral drugs at an early stage of SARS-CoV-2 infection has been a priority for controlling COVID-19. Although Paxlovid and molnupiravir have received emergency approval from the FDA, some side effect concerns have emerged, and the possible oral agents are still limited, resulting in optimized drug development becoming an urgent requirement. An oral remdesivir derivative, VV116, has been reported to have promising antiviral effects against SARS-CoV-2 and positive therapeutic outcomes in clinical trials. However, whether VV116 has broad-spectrum anti-coronavirus activity and potential synergy with other drugs is not clear. Here, we uncovered the broad-spectrum antiviral potency of VV116 against SARS-CoV-2 variants of concern (VOCs), HCoV-OC43, and HCoV-229E in various cell lines. In vitro drug combination screening targeted RdRp and proteinase, highlighting the synergistic effect of VV116 and nirmatrelvir on HCoV-OC43 and SARS-CoV-2. When co-administrated with ritonavir, the combination of VV116 and nirmatrelvir showed significantly enhanced antiviral potency with noninteracting pharmacokinetic properties in mice. Our findings will facilitate clinical treatment with VV116 or VV116+nirmatrelvir combination to fight coronavirus infection.

Abstract Drug development based on target proteins has been a successful approach in recent decades. However, the conventional structure-based drug design (SBDD) pipeline is a complex, human-engineered process with multiple independently optimized steps. Here, we propose a sequence-to-drug concept for computational drug design based on protein sequence information by end-to-end differentiable learning. We validate this concept in three stages. First, we design TransformerCPI2.0 as a core tool for the concept, which demonstrates generalization ability across proteins and compounds. Second, we interpret the binding knowledge that TransformerCPI2.0 learned. Finally, we use TransformerCPI2.0 to discover new hits for challenging drug targets, and identify new target for an existing drug based on an inverse application of the concept. Overall, this proof-of-concept study shows that the sequence-to-drug concept adds a perspective on drug design. It can serve as an alternative method to SBDD, particularly for proteins that do not yet have high-quality 3D structures available.

Colorectal cancer (CRC) ranks as the third most prevalent global malignancy, marked by significant metastasis and post-surgical recurrence, posing formidable challenges to treatment efficacy. The integration of oligonucleotides with chemotherapeutic drugs emerges as a promising strategy for synergistic CRC therapy. The nanoformulation, lipid nanoparticle (LNP), presents the capability to achieve co-delivery of oligonucleotides and chemotherapeutic drugs for cancer therapy. In this study, we constructed lipid nanoparticles, termed as LNP-I-V by microfluidics to co-deliver oligonucleotides miR159 mimics (VDX05001SI) and irinotecan (IRT), demonstrating effective treatment of CRC both in vitro and in vivo. The LNP-I-V exhibited a particle size of 118.67 ± 1.27 nm, ensuring excellent stability and targeting delivery to tumor tissues, where it was internalized and escaped from the endosome with a pH-sensitive profile. Ultimately, LNP-I-V significantly inhibited CRC growth, extended the survival of tumor-bearing mice, and displayed favorable safety profiles. Thus, LNP-I-V held promise as an innovative platform to combine gene therapy and chemotherapy for improving CRC treatment.

Abstract Artificial intelligence (AI)-based molecular design methods, especially deep generative models for generating novel molecule structures, have gratified our imagination to explore unknown chemical space without relying on brute-force exploration. However, whether designed by AI or human experts, the molecules need to be accessibly synthesized and biologically evaluated, and the trial-and-error process remains a resources-intensive endeavor. Therefore, AI-based drug design methods face a major challenge of how to prioritize the molecular structures with potential for subsequent drug development. This study indicates that common filtering approaches based on traditional screening metrics fail to differentiate AI-designed molecules. To address this issue, we propose a novel molecular filtering method, MolFilterGAN, based on a progressively augmented generative adversarial network. Comparative analysis shows that MolFilterGAN outperforms conventional screening approaches based on drug-likeness or synthetic ability metrics. Retrospective analysis of AI-designed discoidin domain receptor 1 (DDR1) inhibitors shows that MolFilterGAN significantly increases the efficiency of molecular triaging. Further evaluation of MolFilterGAN on eight external ligand sets suggests that MolFilterGAN is useful in triaging or enriching bioactive compounds across a wide range of target types. These results highlighted the importance of MolFilterGAN in evaluating molecules integrally and further accelerating molecular discovery especially combined with advanced AI generative models.

As a major class of biomolecules, carbohydrates play indispensable roles in various biological processes. However, it remains largely unknown how carbohydrates directly modulate important drug targets, such as G-protein coupled receptors (GPCRs). Here, we employed P2Y purinoceptor 14 (P2Y14), a drug target for inflammation and immune responses, to uncover the sugar nucleotide activation of GPCRs. Integrating molecular dynamics simulation with functional study, we identified the uridine diphosphate (UDP)-sugar-binding site on P2Y14, and revealed that a UDP-glucose might activate the receptor by bridging the transmembrane (TM) helices 2 and 7. Between TM2 and TM7 of P2Y14, a conserved salt bridging chain (K2.60-D2.64-K7.35-E7.36 [KDKE]) was identified to distinguish different UDP-sugars, including UDP-glucose, UDP-galactose, UDP-glucuronic acid, and UDP-N-acetylglucosamine. We identified the KDKE chain as a conserved functional motif of sugar binding for both P2Y14 and P2Y purinoceptor 12 (P2Y12), and then designed three sugar nucleotides as agonists of P2Y12. These results not only expand our understanding for activation of purinergic receptors but also provide insights for the carbohydrate drug development for GPCRs.

Endothelin receptors (ETAR and ETBR) are critical for vasoregulation and are targets for cardiovascular diseases treatment. Here, the authors offer a structural basis for peptide recognition selectivity and activation of both endothelin receptors.

Follicle stimulating hormone (FSH) is a glycoprotein hormone the functions of which are mediated by a G protein-coupled receptor, FSHR. Here, Duan et al. report cryo-EM structures of FSHR in active and inactive states, suggesting the molecular basis of FSH and small allosteric agonist-mediated FSHR activation.

Summary: In castration-resistant prostate cancer (CRPC), clinical response to androgen receptor (AR) antagonists is limited mainly due to AR-variants expression and restored AR signaling. The metabolite spermine is most abundant in prostate and it decreases as prostate cancer progresses, but its functions remain poorly understood. Here, we show spermine inhibits full-length androgen receptor (AR-FL) and androgen receptor splice variant 7 (AR-V7) signaling and suppresses CRPC cell proliferation by directly binding and inhibiting protein arginine methyltransferase PRMT1. Spermine reduces H4R3me2a modification at the AR locus and suppresses AR binding as well as H3K27ac modification levels at AR target genes. Spermine supplementation restrains CRPC growth in vivo. PRMT1 inhibition also suppresses AR-FL and AR-V7 signaling and reduces CRPC growth. Collectively, we demonstrate spermine as an anticancer metabolite by inhibiting PRMT1 to transcriptionally inhibit AR-FL and AR-V7 signaling in CRPC, and we indicate spermine and PRMT1 inhibition as powerful strategies overcoming limitations of current AR-based therapies in CRPC.

Abstract Endogenous ions play important roles in the function and pharmacology of G protein-coupled receptors (GPCRs) with limited atomic evidence. In addition, compared with G protein subtypes Gs, Gi/o, and Gq/11, insufficient structural evidence is accessible to understand the coupling mechanism of G12/13 protein by GPCRs. Orphan receptor GPR35, which is predominantly expressed in the gastrointestinal tract and is closely related to inflammatory bowel diseases (IBDs), stands out as a prototypical receptor for investigating ionic modulation and G13 coupling. Here we report a cryo-electron microscopy structure of G13-coupled GPR35 bound to an anti-allergic drug, lodoxamide. This structure reveals a novel divalent cation coordination site and a unique ionic regulatory mode of GPR35 and also presents a highly positively charged binding pocket and the complementary electrostatic ligand recognition mode, which explain the promiscuity of acidic ligand binding by GPR35. Structural comparison of the GPR35–G13 complex with other G protein subtypes-coupled GPCRs reveals a notable movement of the C-terminus of α5 helix of the Gα13 subunit towards the receptor core and the least outward displacement of the cytoplasmic end of GPR35 TM6. A featured ‘methionine pocket’ contributes to the G13 coupling by GPR35. Together, our findings provide a structural basis for divalent cation modulation, ligand recognition, and subsequent G13 protein coupling of GPR35 and offer a new opportunity for designing GPR35-targeted drugs for the treatment of IBDs.

Glycolytic metabolism enzymes have been implicated in the immunometabolism field through changes in metabolic status. PGK1 is a catalytic enzyme in the glycolytic pathway. Here, we set up a high-throughput screen platform to identify PGK1 inhibitors. DC-PGKI is an ATP-competitive inhibitor of PGK1 with an affinity of Kd = 99.08 nmol/L. DC-PGKI stabilizes PGK1 in vitro and in vivo, and suppresses both glycolytic activity and the kinase function of PGK1. In addition, DC-PGKI unveils that PGK1 regulates production of IL-1β and IL-6 in LPS-stimulated macrophages. Mechanistically, inhibition of PGK1 with DC-PGKI results in NRF2 (nuclear factor-erythroid factor 2-related factor 2, NFE2L2) accumulation, then NRF2 translocates to the nucleus and binds to the proximity region of Il-1β and Il-6 genes, and inhibits LPS-induced expression of these genes. DC-PGKI ameliorates colitis in the dextran sulfate sodium (DSS)-induced colitis mouse model. These data support PGK1 as a regulator of macrophages and suggest potential utility of PGK1 inhibitors in the treatment of inflammatory bowel disease.

Abstract Blood–brain barrier is a pivotal factor to be considered in the process of central nervous system (CNS) drug development, and it is of great significance to rapidly explore the blood–brain barrier permeability (BBBp) of compounds in silico in early drug discovery process. Here, we focus on whether and how uncertainty estimation methods improve in silico BBBp models. We briefly surveyed the current state of in silico BBBp prediction and uncertainty estimation methods of deep learning models, and curated an independent dataset to determine the reliability of the state-of-the-art algorithms. The results exhibit that, despite the comparable performance on BBBp prediction between graph neural networks-based deep learning models and conventional physicochemical-based machine learning models, the GROVER-BBBp model shows greatly improvement when using uncertainty estimations. In particular, the strategy combined Entropy and MC-dropout can increase the accuracy of distinguishing BBB + from BBB − to above 99% by extracting predictions with high confidence level (uncertainty score

Considerable developments have been observed in fragment-based lead/drug discovery (FBLD/FBDD) recently, with four drugs approved and many others under investigation. Nuclear magnetic resonance (NMR) has gained increasing popularity in FBLD due to its intrinsic capability in characterizing protein-ligand interactions in a large dynamic range of affinity, from weak hits to highly potent drugs. Here, we summarize NMR applications in fragment-based hit-to-lead evolution, including the construction of a fragment library, screening methods, spectra processing, and the delineation of the protein-ligand binding modes. These state-of-the-art NMR techniques have been exemplified in the discovery of inhibitors against multiple targets over the past five years, and they are expected to continue to provide new insights in the future.

Monitoring the physiological changes of organelles is essential for understanding the local biological information of cells and for improving the diagnosis and therapy of diseases. Currently, fluorescent probes are considered as the most powerful tools for imaging and have been widely applied in biomedical fields. However, the expected targeting effects of these probes are often inconsistent with the real experiments. The design of fluorescent probes mainly depends on the empirical knowledge of researchers, which was inhibited by limited chemical space and low efficiency. Herein, we proposed a novel multilevel framework for the prediction of organelle-targeted fluorescent probes by employing advanced artificial intelligence algorithms. In this way, not only the targeting mechanism could be interpreted beyond intuitions but also a quick evaluation method could be established for the rational design. Furthermore, the targeting and imaging powers of the optimized and synthesized probes based on this methodology were verified by quantitative calculation and experiments.

In the paper, Dr. Wang et al reported a cryo-EM structure of the human leukotriene B4 receptor 1 (BLT1) in complex with its native ligand leukotriene B4 (LTB4) in an active conformation complexed with Gi protein. The structure reveals the molecule determinant of LTB4 binding and the mechanism of receptor activation. These structural information will boost the understanding of LTB4-BLT1 signaling and provide a rational basis for designing novel anti-leukotriene drugs.

The basis for the diverse peptide-binding modes and the G protein selectivity of peptide GPCRs remains elusive. Here, the authors offer a structural basis for allosteric-like agonism and G protein selectivity of a neuropeptide GPCR, galanin receptor.

Luteinizing hormone-choriogonadotropin receptor (LHCGR), a class A G protein-coupled receptor (GPCR), plays a pivotal role in the maturation of reproductive organs and embryonic development. Compared with other GPCRs, the subfamily of LHCGR has a large extracellular domain (ECD) to interact with glycoprotein hormones. A unique hinge region connects the ECD and transmembrane domain (TMD) to transfer the activation signal. However, the signal transmission mechanism remains largely unknown. Here, both molecular dynamics simulation and evolutional analysis were applied to explore the effect of the hinge region on signal transmission. The glycoprotein hormone determined specific hinge region conformations, including the position of a long hinge loop and the ECD-TMD interface. With the hormone, the hinge region showed a characteristic rotation and displayed an active-like conformational landscape of the ECD-TMD interface with an extended TMD. The active-like hinge region conformation transduces the hormone binding signal downwards from ECD to TMD. The relationship between the hinge region and the intracelluar G protein-binding pocket was also inferred. The hinge region-mediated signal transmission mechanism offers a deeper understanding of LHCGR and provides insights into the elucidation of GPCR activation.

The TEA domain (TEAD) family proteins (TEAD1‒4) are essential transcription factors that control cell differentiation and organ size in the Hippo pathway. Although the sequences and structures of TEAD family proteins are highly conserved, each TEAD isoform has unique physiological and pathological functions. Therefore, the development and discovery of subtype selective inhibitors for TEAD protein will provide important chemical probes for the TEAD-related function studies in development and diseases. Here, we identified a novel TEAD1/3 covalent inhibitor (DC-TEADin1072) with biochemical IC50 values of 0.61 ± 0.02 and 0.58 ± 0.12 μmol/L against TEAD1 and TEAD3, respectively. Further chemical optimization based on DC-TEAD in 1072 yielded a selective TEAD3 inhibitor DC-TEAD3in03 with the IC50 value of 0.16 ± 0.03 μmol/L, which shows 100-fold selectivity over other TEAD isoforms in activity-based protein profiling (ABPP) assays. In cells, DC-TEAD3in03 showed selective inhibitory effect on TEAD3 in GAL4-TEAD (1–4) reporter assays with the IC50 value of 1.15 μmol/L. When administered to zebrafish juveniles, experiments showed that DC-TEAD3in03 reduced the growth rate of zebrafish caudal fins, indicating the importance of TEAD3 activity in controlling proportional growth of vertebrate appendages.

Abstract Reliable uncertainty quantification for statistical models is crucial in various downstream applications, especially for drug design and discovery where mistakes may incur a large amount of cost. This topic has therefore absorbed much attention and a plethora of methods have been proposed over the past years. The approaches that have been reported so far can be mainly categorized into two classes: distance-based approaches and Bayesian approaches. Although these methods have been widely used in many scenarios and shown promising performance with their distinct superiorities, being overconfident on out-of-distribution examples still poses challenges for the deployment of these techniques in real-world applications. In this study we investigated a number of consensus strategies in order to combine both distance-based and Bayesian approaches together with post-hoc calibration for improved uncertainty quantification in QSAR (Quantitative Structure–Activity Relationship) regression modeling. We employed a set of criteria to quantitatively assess the ranking and calibration ability of these models. Experiments based on 24 bioactivity datasets were designed to make critical comparison between the model we proposed and other well-studied baseline models. Our findings indicate that the hybrid framework proposed by us can robustly enhance the model ability of ranking absolute errors. Together with post-hoc calibration on the validation set, we show that well-calibrated uncertainty quantification results can be obtained in domain shift settings. The complementarity between different methods is also conceptually analyzed.

SARS-CoV-2 3CL protease (3CLpro) is essential for coronavirus replication and of great interest as an antiviral drug target. Here, the authors show that the naturally occurring flavonoid myricetin is a non-peptidomimetic and covalent inhibitor of 3CLpro, and they solve crystal structures of 3CLpro with myricetin and derivatives, which reveal that the pyrogallol group covalently modifies the catalytic cysteine.

Simultaneously visualizing Amyloid-β (Aβ) plaque with its surrounding brain structures at the subcellular level in the intact brain is essential for understanding the complex pathology of Alzheimer's disease, but is still rarely achieved due to the technical limitations. Combining the micro-optical sectioning tomography (MOST) system, whole-brain Nissl staining, and customized image processing workflow, we generated a whole-brain panorama of Alzheimer's disease mice without specific labeling. The workflow employed the steps that include virtual channel splitting, feature enhancement, iso-surface rendering, direct volume rendering, and feature fusion to extract and reconstruct the different signals with distinct gray values and morphologies. Taking advantage of this workflow, we found that the denser-distribution areas of Aβ plaques appeared with relatively more somata and smaller vessels, but show a dissimilar distributing pattern with nerve tracts. In addition, the entorhinal cortex and adjacent subiculum regions present the highest density and biggest diameter of plaques. The neuronal processes in the vicinity of these Aβ plaques showed significant structural alternation such as bending or abrupt branch ending. The capillaries inside or adjacent to the plaques were observed with abundant distorted micro-vessels and abrupt ending. Depicting Aβ plaques, somata, nerve processes and tracts, and blood vessels simultaneously, this panorama enables us for the first time, to analyze how the Aβ plaques interact with capillaries, somata, and processes at a submicron resolution of 3D whole-brain scale, which reveals potential pathological effects of Aβ plaques from a new cross-scale view. Our approach opens a door to routine systematic studies of complex interactions among brain components in mouse models of Alzheimer's disease.

A newly emerged coronavirus, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), belongs to the β-coronavirus family and shows high similarities with SARS-CoV. On March 11, 2020, the World Health Organization (WHO) declared SARS-CoV-2 a global pandemic, and the disease was named the coronavirus disease 2019 (COVID-19). The ongoing COVID-19 pandemic has caused over 46 million infections and over one million deaths worldwide, and the numbers are still increasing. Efficacious antiviral agents are urgently needed to combat this virus. The life cycle of SARS-CoV-2 mainly includes the viral attachment, membrane fusion, genomic replication, assembly and budding of virions. Accordingly, drug development against SARS-CoV-2 currently focuses on blocking spike protein binding to ACE2, inhibiting viral membrane fusion with host cells, and preventing the viral replication by targeting 3C-like protease, papain-like protease, RNA-dependent RNA polymerase as well as some host-cell proteins. In this review, the advances of drug development in these three major areas are elaborated.

Fibroblast growth factor receptors (FGFRs) have emerged as promising targets for anticancer therapy. In this study, we synthesized and evaluated the biological activity of 66 pyrazolo[3,4-d]pyridazinone derivatives. Kinase inhibition, cell proliferation, and whole blood stability assays were used to evaluate their activity on FGFR, allowing us to explore structure−activity relationships and thus to gain understanding of the structural requirements to modulate covalent inhibitors’ selectivity and reactivity. Among them, compound 10h exhibited potent enzymatic activity against FGFR and remarkably inhibited proliferation of various cancer cells associated with FGFR dysregulation, and suppressed FGFR signaling pathway in cancer cells by the immunoblot analysis. Moreover, 10h displayed highly potent antitumor efficacy (TGI = 91.6%, at a dose of 50 mg/kg) in the FGFR1-amplified NCI-H1581 xenograft model.

Growth hormone-releasing hormone (GHRH) controls metabolism and tissue growth through binding to the cognate receptor (GHRHR). Here authors report the structure of the human GHRHR bound to its endogenous ligand and the stimulatory G protein which reveals a characteristic hormone recognition pattern of GHRH by GHRHR.

Glucagon-like peptide-1 receptor (GLP-1R) plays an important role in glucose homeostasis and treatment of type 2 diabetes. Here authors report the peptide-free crystal structure of human GLP-1R in an inactive state which reveals a unique closed conformation of the extracellular domain.

Proprotein convertase subtilisin/kexin type 9 (PCSK9) modulators may attenuate PCSK9-induced low-density lipoprotein receptor (LDLR) degradation in lysosome and promote the clearance of circulating low-density lipoprotein cholesterol (LDL-C). A novel series of tetrahydroprotoberberine derivatives (THPBs) were designed, synthesized, and evaluated as PCSK9 modulators for the treatment of hyperlipidemia. Among them, eight compounds exhibited excellent activities in downregulating hepatic PCSK9 expression better than berberine in HepG2 cells. In addition, five compounds 15, 18, 22, (R)-22, and (S)-22 showed better performance in the low-density lipoprotein, labeled with 1,1′-dioctadecyl-3,3,3′,3′-tetramethyl-indocarbocyanine perchlorate (DiI-LDL) uptake assay, compared with berberine at the same concentration. Compound 22, selected for in vivo evaluation, demonstrated significant reductions of total cholesterol (TC) and LDL-C in hyperlipidemic hamsters with a good pharmacokinetic profile. Further exploring of the lipid-lowering mechanism showed that compound 22 promoted hepatic LDLR expression in a dose-dependent manner in HepG2 cells. Additional results of human ether-à-go-go related gene (hERG) inhibition assay indicated the potential druggability for compound 22, which is a promising lead compound for the development of PCSK9 modulator for the treatment of hyperlipidemia. Key words: PCSK9, Tetrahydroprotoberberine derivatives, Low-density lipoprotein cholesterol, Lipid-lowering, PCSK9 expression, Low-density lipoprotein receptor, Total cholesterol, Hyperlipidemia hamster

Deep learning is very promising in solving problems in omics research, such as genomics, epigenomics, proteomics, and metabolics. The design of neural network architecture is very important in modeling omics data against different scientific problems. Residual fully-connected neural network (RFCN) was proposed to provide better neural network architectures for modeling omics data. The next challenge for omics research is how to integrate information from different omics data using deep learning, so that information from different molecular system levels could be combined to predict the target. In this paper, we present a novel multi-omics integration approach named AutoOmics that could efficiently integrate information from different omics data and achieve better accuracy than previous approaches. We evaluated our method on four different tasks: drug repositioning, target gene prediction, breast cancer subtyping and cancer type prediction, and all the four tasks achieved state of art performances.

Omics data can be used to identify biological characteristics from genetic to phenotypic levels during the life span of a living being, while molecular interaction networks have a fundamental impact on life activities. Integrating omics data and molecular interaction networks will help researchers delve into comprehensive information hidden in the data. Here, we propose a new multimodal method — AutoGGN — to integrate multi-omics data with molecular interaction networks based on graph convolutional neural networks (GCNs). We evaluated AutoGGN using three classification tasks: single-cell embryonic developmental stage classification, pan-cancer type classification, and breast cancer subtyping. On all three tasks, AutoGGN showed better performance than other methods. This means AutoGGN has the potential to extract insights more effectively by means of integrating molecular interaction networks with multi-omics data. Additionally, in order to provide a better understanding of how our model makes predictions, we utilized the SHAP module and identified the key genes contributing to the classification, providing insight for the design of downstream biological experiments.

A continuing challenge in modern medicine is the identification of safer and more efficacious drugs. Precision therapeutics, which have one molecular target, have been long promised to be safer and more effective than traditional therapies. This approach has proven to be challenging for multiple reasons including lack of efficacy, rapidly acquired drug resistance, and narrow patient eligibility criteria. An alternative approach is the development of drugs that address the overall disease network by targeting multiple biological targets ('polypharmacology'). Rational development of these molecules will require improved methods for predicting single chemical structures that target multiple drug targets. To address this need, we developed the Multi-Targeting Drug DREAM Challenge, in which we challenged participants to predict single chemical entities that target pro-targets but avoid anti-targets for two unrelated diseases: RET-based tumors and a common form of inherited Tauopathy. Here, we report the results of this DREAM Challenge and the development of two neural network-based machine learning approaches that were applied to the challenge of rational polypharmacology. Together, these platforms provide a potentially useful first step towards developing lead therapeutic compounds that address disease complexity through rational polypharmacology.

Abstract The effective pulmonary deposition of inhaled particulate carriers loaded with drugs is a prerequisite for therapeutic effects of drug delivery via inhalation route. Revealing the sophisticated lung scaffold and intrapulmonary distribution of particles at three‐dimensional (3D), in‐situ, and single‐particle level remains a fundamental and critical challenge for dry powder inhalation in pre‐clinical research. Here, taking advantage of the micro optical sectioning tomography system, the high‐precision cross‐scale visualization of entire lung anatomy is obtained. Then, co‐localized lung‐wide datasets of both cyto‐architectures and fluorescent particles are collected at full scale with the resolution down to individual particles. The precise spatial distribution pattern reveals the region‐specific distribution and structure‐associated deposition of the inhalable particles in lungs, which is undetected by previous methods. Overall, this research delivers comprehensive and high‐resolution 3D detection of pulmonary drug delivery vectors and provides a novel strategy to evaluate materials distribution for drug delivery.

Abstract The Rho family GTPases are crucial drivers of tumor growth and metastasis. However, it is difficult to develop GTPases inhibitors due to a lack of well‐characterized binding pockets for compounds. Here, through molecular dynamics simulation of the RhoA protein, a groove around cysteine 107 (Cys107) that is relatively well‐conserved within the Rho family is discovered. Using a combined strategy, the novel inhibitor DC‐Rhoin is discovered, which disrupts interaction of Rho proteins with guanine nucleotide exchange factors (GEFs) and guanine nucleotide dissociation inhibitors (GDIs). Crystallographic studies reveal that the covalent binding of DC‐Rhoin to the Cys107 residue stabilizes and captures a novel allosteric pocket. Moreover, the derivative compound DC‐Rhoin04 inhibits the migration and invasion of cancer cells, through targeting this allosteric pocket of RhoA. The study reveals a novel allosteric regulatory site within the Rho family, which can be exploited for anti‐metastasis drug development, and also provides a novel strategy for inhibitor discovery toward “undruggable” protein targets.

Summary: The application of machine learning toward DNA encoded library (DEL) technology is lacking despite obvious synergy between these two advancing technologies. Herein, a machine learning algorithm has been developed that predicts the conversion rate for the DNA-compatible reaction of a building block with a model DNA-conjugate. We exemplify the value of this technique with a challenging reaction, the Pictet-Spengler, where acidic conditions are normally required to achieve the desired cyclization between tryptophan and aldehydes to provide tryptolines. This is the first demonstration of using a machine learning algorithm to cull potential building blocks prior to their purchase and testing for DNA-encoded library synthesis. Importantly, this allows for a challenging reaction, with an otherwise very low building block pass rate in the test reaction, to still be used in DEL synthesis. Furthermore, because our protocol is solution phase it is directly applicable to standard plate-based DEL synthesis.

Glycosylation is an important post-translational modification that affects a wide variety of physiological functions. DC-SIGN (Dendritic Cell-Specific Intercellular adhesion molecule-3-Grabbing Non-integrin) is a protein expressed in antigen-presenting cells that recognizes a variety of glycan epitopes. Until now, the binding of DC-SIGN to SARS-CoV-2 Spike glycoprotein has been reported in various articles and is regarded to be a factor in systemic infection and cytokine storm. The mechanism of DC-SIGN recognition offers an alternative method for discovering new medication for COVID-19 treatment. Here, we discovered three potential pockets that hold different glycan epitopes by performing molecular dynamics simulations of previously reported oligosaccharides. The “EPN” motif, “NDD” motif, and Glu354 form the most critical pocket, which is known as the Core site. We proposed that the type of glycan epitopes, rather than the precise amino acid sequence, determines the recognition. Furthermore, we deduced that oligosaccharides could occupy an additional site, which adds to their higher affinity than monosaccharides. Based on our findings and previously described glycoforms on the SARS-CoV-2 Spike, we predicted the potential glycan epitopes for DC-SIGN. It suggested that glycan epitopes could be recognized at multiple sites, not just Asn234, Asn149 and Asn343. Subsequently, we found that Saikosaponin A and Liquiritin, two plant glycosides, were promising DC-SIGN antagonists in silico.

Scoring functions play an important role in structure-based virtual screening. It has been widely accepted that target-specific scoring functions (TSSFs) may achieve better performance compared with universal scoring functions in actual drug research and development processes. A method that can effectively construct TSSFs will be of great value to drug design and discovery. In this work, we proposed a deep learning–based model named DeepScore to achieve this goal. DeepScore adopted the form of PMF scoring function to calculate protein–ligand binding affinity. However, different from PMF scoring function, in DeepScore, the score for each protein–ligand atom pair was calculated using a feedforward neural network. Our model significantly outperformed Glide Gscore on validation data set DUD-E. The average ROC-AUC on 102 targets was 0.98. We also combined Gscore and DeepScore together using a consensus method and put forward a consensus model named DeepScoreCS. The comparison results showed that DeepScore outperformed other machine learning–based TSSFs building methods. Furthermore, we presented a strategy to visualize the prediction of DeepScore. All of these results clearly demonstrated that DeepScore would be a useful model in constructing TSSFs and represented a novel way incorporating deep learning and drug design.

Elucidating the conformational heterogeneity of proteins is essential for understanding protein function and developing exogenous ligands. With the rapid development of experimental and computational methods, it is of great interest to integrate these approaches to illuminate the conformational landscapes of target proteins. SETD8 is a protein lysine methyltransferase (PKMT), which functions in vivo via the methylation of histone and nonhistone targets. Utilizing covalent inhibitors and depleting native ligands to trap hidden conformational states, we obtained diverse X-ray structures of SETD8. These structures were used to seed distributed atomistic molecular dynamics simulations that generated a total of six milliseconds of trajectory data. Markov state models, built via an automated machine learning approach and corroborated experimentally, reveal how slow conformational motions and conformational states are relevant to catalysis. These findings provide molecular insight on enzymatic catalysis and allosteric mechanisms of a PKMT via its detailed conformational landscape.

The WalKR signal transduction system is involved in extracellular signal recognition, but the details of this function are not well established. Here, the authors report the crystal structure of this two-component system alongside the characterisation of a small-molecule activator.

Human 5-lipoxygenase (5-LOX) is a well-validated drug target and its inhibitors are potential drugs for treating leukotriene-related disorders. Our previous work on structural optimization of the hit compound 2 from our in-house collection identified two lead compounds, 3a and 3b, exhibiting a potent inhibitory profile against 5-LOX with IC50 values less than 1 µmol/L in cell-based assays. Here, we further optimized these compounds to prepare a class of novel pyrazole derivatives by opening the fused-ring system. Several new compounds exhibited more potent inhibitory activity than the lead compounds against 5-LOX. In particular, compound 4e not only suppressed lipopolysaccharide-induced inflammation in brain inflammatory cells and protected neurons from oxidative toxicity, but also significantly decreased infarct damage in a mouse model of cerebral ischemia. Molecular docking analysis further confirmed the consistency of our theoretical results and experimental data. In conclusion, the excellent in vitro and in vivo inhibitory activities of these compounds against 5-LOX suggested that these novel chemical structures have a promising therapeutic potential to treat leukotriene-related disorders.

Ion channels are important therapeutic targets, and their pharmacology is becoming increasingly important. However, knowledge of the mechanism of interaction of the activators and ion channels is still limited due to the complexity of the mechanisms. A statistical thermodynamic model has been developed in this study to characterize the cooperative binding of activators to ion channels. By fitting experimental concentration-response data, the model gives eight parameters for revealing the mechanism of an activator potentiating an ion channel, i.e., the binding affinity (KA), the binding cooperative coefficients for two to four activator molecules interacting with one channel (γ, μ, and ν), and the channel conductance coefficients for four activator binding configurations of the channel (a, b, c, and d). Values for the model parameters and the mechanism underlying the interaction of ztz240, a proven KCNQ2 activator, with the wild-type channel have been obtained and revealed by fitting the concentration-response data of this activator potentiating the outward current amplitudes of KCNQ2. With these parameters, our model predicted an unexpected bi-sigmoid concentration-response curve of ztz240 activation of the WT-F137A mutant heteromeric channel that was in good agreement with the experimental data determined in parallel in this study, lending credence to the assumptions on which the model is based and to the model itself. Our model can provide a better fit to the measured data than the Hill equation and estimates the binding affinity, as well as the cooperative coefficients for the binding of activators and conductance coefficients for binding states, which validates its use in studying ligand-channel interaction mechanisms.

In this work, the relationship between cyclophilin A (CypA) and HCV prompted us to screen a series of small molecule CypA inhibitors which were previously reported by our group. Among them, compound 1, discovered as a non-immunosuppressive anti-HCV agent with an EC50 value of 0.67 μM in a virus assay, was selected for further study. Subsequent chemical modification by O-acylation led to a novel class of molecules, among which compound 25 demonstrated the most potent anti-HCV activity in the virus assay (EC50 = 0.19 μM), but low cytotoxicity and hERG cardiac toxicity. The following studies (a solution stability assay and a simple pharmacokinetic test together with a CypA enzyme inhibition assay) preliminarily indicated that 25 was a prodrug of 1. To the best of our knowledge, 25 is probably the most potent currently reported small molecule anti-HCV agent acting via the CypA inhibitory mechanism. Consequently, our study has provided a new potential small molecule for curing HCV infection.

The fibroblast growth factor/fibroblast growth factor receptor (FGF/FGFR) signaling pathway plays crucial roles in cell proliferation, angiogenesis, migration, and survival. Aberration in FGFRs correlates with several malignancies and disorders. FGFRs have proved to be attractive targets for therapeutic intervention in cancer, and it is of high interest to find FGFR inhibitors with novel scaffolds. In this study, a combinatorial three-dimensional quantitative structure-activity relationship (3D-QSAR) model was developed based on previously reported FGFR1 inhibitors with diverse structural skeletons. This model was evaluated for its prediction performance on a diverse test set containing 232 FGFR inhibitors, and it yielded a SD value of 0.75 pIC50 units from measured inhibition affinities and a Pearson’s correlation coefficient R2 of 0.53. This result suggests that the combinatorial 3D-QSAR model could be used to search for new FGFR1 hit structures and predict their potential activity. To further evaluate the performance of the model, a decoy set validation was used to measure the efficiency of the model by calculating EF (enrichment factor). Based on the combinatorial pharmacophore model, a virtual screening against SPECS database was performed. Nineteen novel active compounds were successfully identified, which provide new chemical starting points for further structural optimization of FGFR1 inhibitors.

Hepatitis C virus (HCV) NS3/NS4A serine protease is essential for viral replication, which is regarded as a promising drug target for developing direct-acting anti-HCV agents. In this study, sixteen novel compounds with cell-based HCV replicon activity ranging from 3.0 to 28.2 μM (IC50) were successfully identified by means of structure-based virtual screening. Compound 5 and compound 11, with an IC50 of 3.0 μM and 5.1 μM, respectively, are the two most potent molecules with low cytotoxicity.

An efficient approach for the synthesis of biologically interesting fused tetracyclic isoquinolines in high yields and with a broad substrate scope has been developed. The strategy features an AgNO3 catalyzed ‘one-pot’ cascade process involving formation of two new C–N bonds and one new C–O bond.

An efficient one-pot method to generate structurally diverse and medicinally interesting pyrazolone derivatives in good to excellent yields of 51–98% under microwave irradiation and solvent-free conditions has been developed.