H2A.Z facilitates licensing and activation of early replication origins.

DNA replication is a tightly regulated process that ensures the precise duplication of the genome during the cell cycle1. In eukaryotes, the licensing and activation of replication origins are regulated by both DNA sequence and chromatin features2. However, the chromatin-based regulatory mechanisms remain largely uncharacterized. Here we show that, in HeLa cells, nucleosomes containing the histone variant H2A.Z are enriched with histone H4 that is dimethylated on its lysine 20 residue (H4K20me2) and with bound origin-recognition complex (ORC). In vitro studies show that H2A.Z-containing nucleosomes bind directly to the histone lysine methyltransferase enzyme SUV420H1, promoting H4K20me2 deposition, which is in turn required for ORC1 binding. Genome-wide studies show that signals from H4K20me2, ORC1 and nascent DNA strands co-localize with H2A.Z, and that depletion of H2A.Z results in decreased H4K20me2, ORC1 and nascent-strand signals throughout the genome. H2A.Z-regulated replication origins have a higher firing efficiency and early replication timing compared with other origins. Our results suggest that the histone variant H2A.Z epigenetically regulates the licensing and activation of early replication origins and maintains replication timing through the SUV420H1-H4K20me2-ORC1 axis.

Author Correction: H2A.Z facilitates licensing and activation of early replication origins.

An Amendment to this paper has been published and can be accessed via a link at the top of the paper.

Challenging Complexity with Simplicity: Rethinking the Role of Single-Step Models in Computer-Aided Synthesis Planning.

Computer-assisted synthesis planning has become increasingly important in drug discovery. While deep-learning models have shown remarkable progress in achieving high accuracies for single-step retrosynthetic predictions, their performances in retrosynthetic route planning need to be checked. This study compares the intricate single-step models with a straightforward template enumeration approach for retrosynthetic route planning on a real-world drug molecule data set. Despite the superior single-step accuracy of advanced models, the template enumeration method with a heuristic-based retrosynthesis knowledge score was found to surpass them in efficiency in searching the reaction space, achieving a higher or comparable solve rate within the same time frame. This counterintuitive result underscores the importance of efficiency and retrosynthesis knowledge in retrosynthesis route planning and suggests that future research should incorporate a simple template enumeration as a benchmark. It also suggests that this simple yet effective strategy should be considered alongside more complex models to better cater to the practical needs of computer-assisted synthesis planning in drug discovery.

A transferable deep learning approach to fast screen potential antiviral drugs against SARS-CoV-2.

The COVID-19 pandemic calls for rapid development of effective treatments. Although various drug repurpose approaches have been used to screen the FDA-approved drugs and drug candidates in clinical phases against SARS-CoV-2, the coronavirus that causes this disease, no magic bullets have been found until now. In this study, we used directed message passing neural network to first build a broad-spectrum anti-beta-coronavirus compound prediction model, which gave satisfactory predictions on newly reported active compounds against SARS-CoV-2. Then, we applied transfer learning to fine-tune the model with the recently reported anti-SARS-CoV-2 compounds and derived a SARS-CoV-2 specific prediction model COVIDVS-3. We used COVIDVS-3 to screen a large compound library with 4.9 million drug-like molecules from ZINC15 database and recommended a list of potential anti-SARS-CoV-2 compounds for further experimental testing. As a proof-of-concept, we experimentally tested seven high-scored compounds that also demonstrated good binding strength in docking studies against the 3C-like protease of SARS-CoV-2 and found one novel compound that can inhibit the enzyme. Our model is highly efficient and can be used to screen large compound databases with millions or more compounds to accelerate the drug discovery process for the treatment of COVID-19.

Anxiety prevalence and its association with physical activity in patients with non-communicable diseases during COVID-19 lockdown: a cross-sectional study in Shanghai, China.

BACKGROUND: Quarantine due to the COVID-19 pandemic may have created great psychological stress among vulnerable populations. We aimed to investigate the prevalence of anxiety and explore the association between physical activities (PA) and anxiety risk in people with non-communicable diseases during the period of COVID-19 lockdown. METHODS: We conducted a cross-sectional telephone survey from February 25 to April 20, 2020, the period of COVID-19 lockdown in Shanghai. Up to 8000 patients with type 2 diabetes and/or hypertension were selected using multi-stage cluster random sampling. PA level was measured based on the International Physical Activity Questionnaire using Metabolic Equivalent for Task scores, while symptoms of anxiety were assessed by the 7-item Generalized Anxiety Disorder scale. Multiple logistic regression analyses were performed to evaluate the associations of type and level of PA with the risk of anxiety. RESULTS: Of a total 4877 eligible patients, 2602 (53.4%) reported with anxiety, and 2463 (50.5%), 123 (2.5%) and 16 (0.3%) reported with mild, moderate, and severe anxiety. The prevalence of anxiety was higher in the females, the elders, non-smokers, non-drinkers, and patients with diabetes, and the associations of anxiety with sex, age, smoking, drinking and diagnosis of diabetes were significant. A significant negative association was observed for housework activities (OR 0.53, 95%CI: [0.45, 0.63], p < 0.001) and trip activities (OR 0.55, 95%CI: [0.48, 0.63], p < 0.001) with anxiety, but no significant was found for exercise activities (OR 1.06, 95%CI: [0.94, 1.20], p = 0.321). Compared with patients with a low PA level, those with a moderate (OR 0.53, 95%CI: [0.44, 0.64], p < 0.001) or a high PA level (OR 0.51, 95%CI: [0.43, 0.51], p < 0.001) had a lower prevalence of anxiety. CONCLUSION: This study demonstrates a higher prevalence of anxiety in patients with hypertension, diabetes, or both during the COVID-19 lockdown. The negative associations of housework and trip activities with anxiety highlight the potential benefit of PA among patients with non-communicable diseases.

Prediction of liquid-liquid phase separating proteins using machine learning.

BACKGROUND: The liquid-liquid phase separation (LLPS) of biomolecules in cell underpins the formation of membraneless organelles, which are the condensates of protein, nucleic acid, or both, and play critical roles in cellular function. Dysregulation of LLPS is implicated in a number of diseases. Although the LLPS of biomolecules has been investigated intensively in recent years, the knowledge of the prevalence and distribution of phase separation proteins (PSPs) is still lag behind. Development of computational methods to predict PSPs is therefore of great importance for comprehensive understanding of the biological function of LLPS. RESULTS: Based on the PSPs collected in LLPSDB, we developed a sequence-based prediction tool for LLPS proteins (PSPredictor), which is an attempt at general purpose of PSP prediction that does not depend on specific protein types. Our method combines the componential and sequential information during the protein embedding stage, and, adopts the machine learning algorithm for final predicting. The proposed method achieves a tenfold cross-validation accuracy of 94.71%, and outperforms previously reported PSPs prediction tools. For further applications, we built a user-friendly PSPredictor web server ( http://www.pkumdl.cn/PSPredictor ), which is accessible for prediction of potential PSPs. CONCLUSIONS: PSPredictor could identifie novel scaffold proteins for stress granules and predict PSPs candidates in the human genome for further study. For further applications, we built a user-friendly PSPredictor web server ( http://www.pkumdl.cn/PSPredictor ), which provides valuable information for potential PSPs recognition.

Advances and Challenges in De Novo Drug Design Using Three-Dimensional Deep Generative Models.

A persistent goal for de novo drug design is to generate novel chemical compounds with desirable properties in a labor-, time-, and cost-efficient manner. Deep generative models provide alternative routes to this goal. Numerous model architectures and optimization strategies have been explored in recent years, most of which have been developed to generate two-dimensional molecular structures. Some generative models aiming at three-dimensional (3D) molecule generation have also been proposed, gaining attention for their unique advantages and potential to directly design drug-like molecules in a target-conditioning manner. This review highlights current developments in 3D molecular generative models combined with deep learning and discusses future directions for de novo drug design.

MolMiner: You Only Look Once for Chemical Structure Recognition.

Molecular structures are commonly depicted in 2D printed forms in scientific documents such as journal papers and patents. However, these 2D depictions are not machine readable. Due to a backlog of decades and an increasing amount of printed literatures, there is a high demand for translating printed depictions into machine-readable formats, which is known as Optical Chemical Structure Recognition (OCSR). Most OCSR systems developed over the last three decades use a rule-based approach, which vectorizes the depiction based on the interpretation of vectors and nodes as bonds and atoms. Here, we present a practical software called MolMiner, which is primarily built using deep neural networks originally developed for semantic segmentation and object detection to recognize atom and bond elements from documents. These recognized elements can be easily connected as a molecular graph with a distance-based construction algorithm. MolMiner gave state-of-the-art performance on four benchmark data sets and a self-collected external data set from scientific papers. As MolMiner performed similarly well in real-world OCSR tasks with a user-friendly interface, it is a useful and valuable tool for daily applications. The free download links of Mac and Windows versions are available at https://github.com/iipharma/pharmamind-molminer.

COVID-19 outbreak improves attractiveness of medical careers in Chinese senior high school students.

BACKGROUND: The shortage of healthcare workers is becoming a serious global problem. The underlying reasons may be specific to the healthcare system in each country. Over the past decade, medicine has become an increasingly unpopular profession in China due to the heavy workload, long-term training, and inherent risks. The ongoing COVID-19 pandemic has placed the life-saving roles of healthcare professionals under the spotlight. This public health crisis may have a profound impact on career choices in Chinese population. METHODS: We conducted a questionnaire-based online survey among 21,085 senior high school students and 21,009 parents from 24 provinces (or municipalities) of China. We investigated the change of interest in medical study due to the outbreak of COVID-19 and the potential motivational factors based on the expectancy-value theory framework. Pearson correlation analysis was used to assess the correlation of static or dynamic interest in medical career pursuit with the reported number of COVID-19 cases. Logistic regression model was adopted to analyze the main factors associated with students' choices. RESULTS: We observed an increased preference for medical study post the outbreak of COVID-19 in both students (17.5 to 29.6%) and parents (37.1 to 47.3%). Attainment value was found to be the main reason for the choice among students, with the contribution to society rated as the top motivation. On the other hand, the predominant demotivation in high school students was lack of interest, followed by concerns regarding violence against doctors, heavy workload, long-term training and heavy responsibility as a doctor. Additionally, students who were female, in the resit of final year, had highly educated parents and outside of Hubei province were significantly associated with a keen interest in pursuing medical study. CONCLUSIONS: This is the first multi-center cross-sectional study exploring the positive change and motivations of students' preferences in medical study due to the outbreak of COVID-19. Our results may help medical educators, researchers and policymakers to restructure medical education to make it more appealing to high school students, particularly, to develop a more supportive social and working environment for medical professionals to maintain the observed enhanced enthusiasm.

CavityPlus: a web server for protein cavity detection with pharmacophore modelling, allosteric site identification and covalent ligand binding ability prediction.

CavityPlus is a web server that offers protein cavity detection and various functional analyses. Using protein three-dimensional structural information as the input, CavityPlus applies CAVITY to detect potential binding sites on the surface of a given protein structure and rank them based on ligandability and druggability scores. These potential binding sites can be further analysed using three submodules, CavPharmer, CorrSite, and CovCys. CavPharmer uses a receptor-based pharmacophore modelling program, Pocket, to automatically extract pharmacophore features within cavities. CorrSite identifies potential allosteric ligand-binding sites based on motion correlation analyses between cavities. CovCys automatically detects druggable cysteine residues, which is especially useful to identify novel binding sites for designing covalent allosteric ligands. Overall, CavityPlus provides an integrated platform for analysing comprehensive properties of protein binding cavities. Such analyses are useful for many aspects of drug design and discovery, including target selection and identification, virtual screening, de novo drug design, and allosteric and covalent-binding drug design. The CavityPlus web server is freely available at http://repharma.pku.edu.cn/cavityplus or http://www.pkumdl.cn/cavityplus.

Efficient molecular encoders for virtual screening.

Molecular representations encoding molecular structure information play critical roles in molecular virtual screening (VS). In order to improve VS performance, an abundance of molecular encoders have been developed and tested by various VS challenges. Combinational strategies were also used to improve the performance. Deep learning (DL)-based molecular encoders have attracted much attention for their automatic information extraction ability. In this review, we present an overview of two-dimensional-, three-dimensional-, and DL-based molecular encoders, summarize recent progress of VS using DL technologies, and propose a general framework of DL molecular encoder-based VS. Perspectives on the future directions of molecular representations and applications in the prediction of active compounds are also provided.

PharmMapper 2017 update: a web server for potential drug target identification with a comprehensive target pharmacophore database.

The PharmMapper online tool is a web server for potential drug target identification by reversed pharmacophore matching the query compound against an in-house pharmacophore model database. The original version of PharmMapper includes more than 7000 target pharmacophores derived from complex crystal structures with corresponding protein target annotations. In this article, we present a new version of the PharmMapper web server, of which the backend pharmacophore database is six times larger than the earlier one, with a total of 23 236 proteins covering 16 159 druggable pharmacophore models and 51 431 ligandable pharmacophore models. The expanded target data cover 450 indications and 4800 molecular functions compared to 110 indications and 349 molecular functions in our last update. In addition, the new web server is united with the statistically meaningful ranking of the identified drug targets, which is achieved through the use of standard scores. It also features an improved user interface. The proposed web server is freely available at http://lilab.ecust.edu.cn/pharmmapper/.

Prediction of Human Cytochrome P450 Inhibition Using a Multitask Deep Autoencoder Neural Network.

Adverse side effects of drug-drug interactions induced by human cytochrome P450 (CYP450) inhibition is an important consideration in drug discovery. It is highly desirable to develop computational models that can predict the inhibitive effect of a compound against a specific CYP450 isoform. In this study, we developed a multitask model for concurrent inhibition prediction of five major CYP450 isoforms, namely, 1A2, 2C9, 2C19, 2D6, and 3A4. The model was built by training a multitask autoencoder deep neural network (DNN) on a large dataset containing more than 13 000 compounds, extracted from the PubChem BioAssay Database. We demonstrate that the multitask model gave better prediction results than that of single-task models, previous reported classifiers, and traditional machine learning methods on an average of five prediction tasks. Our multitask DNN model gave average prediction accuracies of 86.4% for the 10-fold cross-validation and 88.7% for the external test datasets. In addition, we built linear regression models to quantify how the other tasks contributed to the prediction difference of a given task between single-task and multitask models, and we explained under what conditions the multitask model will outperform the single-task model, which suggested how to use multitask DNN models more effectively. We applied sensitivity analysis to extract useful knowledge about CYP450 inhibition, which may shed light on the structural features of these isoforms and give hints about how to avoid side effects during drug development. Our models are freely available at http://repharma.pku.edu.cn/deepcyp/home.php or http://www.pkumdl.cn/deepcyp/home.php .

Mutation-induced protein interaction kinetics changes affect apoptotic network dynamic properties and facilitate oncogenesis.

It has been a consensus in cancer research that cancer is a disease caused primarily by genomic alterations, especially somatic mutations. However, the mechanism of mutation-induced oncogenesis is not fully understood. Here, we used the mitochondrial apoptotic pathway as a case study and performed a systematic analysis of integrating pathway dynamics with protein interaction kinetics to quantitatively investigate the causal molecular mechanism of mutation-induced oncogenesis. A mathematical model of the regulatory network was constructed to establish the functional role of dynamic bifurcation in the apoptotic process. The oncogenic mutation enrichment of each of the protein functional domains involved was found strongly correlated with the parameter sensitivity of the bifurcation point. We further dissected the causal mechanism underlying this correlation by evaluating the mutational influence on protein interaction kinetics using molecular dynamics simulation. We analyzed 29 matched mutant-wild-type and 16 matched SNP--wild-type protein systems. We found that the binding kinetics changes reflected by the changes of free energy changes induced by protein interaction mutations, which induce variations in the sensitive parameters of the bifurcation point, were a major cause of apoptosis pathway dysfunction, and mutations involved in sensitive interaction domains show high oncogenic potential. Our analysis provided a molecular basis for connecting protein mutations, protein interaction kinetics, network dynamics properties, and physiological function of a regulatory network. These insights provide a framework for coupling mutation genotype to tumorigenesis phenotype and help elucidate the logic of cancer initiation.

Sequence-based prediction of protein protein interaction using a deep-learning algorithm.

BACKGROUND: Protein-protein interactions (PPIs) are critical for many biological processes. It is therefore important to develop accurate high-throughput methods for identifying PPI to better understand protein function, disease occurrence, and therapy design. Though various computational methods for predicting PPI have been developed, their robustness for prediction with external datasets is unknown. Deep-learning algorithms have achieved successful results in diverse areas, but their effectiveness for PPI prediction has not been tested. RESULTS: We used a stacked autoencoder, a type of deep-learning algorithm, to study the sequence-based PPI prediction. The best model achieved an average accuracy of 97.19% with 10-fold cross-validation. The prediction accuracies for various external datasets ranged from 87.99% to 99.21%, which are superior to those achieved with previous methods. CONCLUSIONS: To our knowledge, this research is the first to apply a deep-learning algorithm to sequence-based PPI prediction, and the results demonstrate its potential in this field.

Computational Multitarget Drug Design.

Designing drugs that can simultaneously interact with multiple targets is a promising approach for treating complicated diseases. Compared to using combinations of single target drugs, multitarget drugs have advantages of higher efficacy, improved safety profile, and simpler administration. Many in silico methods have been developed to approach different aspects of this polypharmacology-guided drug design, particularly for drug repurposing and multitarget drug design. In this review, we summarize recent progress in computational multitarget drug design and discuss their advantages and limitations. Perspectives for future drug development will also be discussed.

Deep Learning Based Regression and Multiclass Models for Acute Oral Toxicity Prediction with Automatic Chemical Feature Extraction.

Median lethal death, LD50, is a general indicator of compound acute oral toxicity (AOT). Various in silico methods were developed for AOT prediction to reduce costs and time. In this study, we developed an improved molecular graph encoding convolutional neural networks (MGE-CNN) architecture to construct three types of high-quality AOT models: regression model (deepAOT-R), multiclassification model (deepAOT-C), and multitask model (deepAOT-CR). These predictive models highly outperformed previously reported models. For the two external data sets containing 1673 (test set I) and 375 (test set II) compounds, the R2 and mean absolute errors (MAEs) of deepAOT-R on the test set I were 0.864 and 0.195, and the prediction accuracies of deepAOT-C were 95.5% and 96.3% on test sets I and II, respectively. The two external prediction accuracies of deepAOT-CR are 95.0% and 94.1%, while the R2 and MAE are 0.861 and 0.204 for test set I, respectively. We then performed forward and backward exploration of deepAOT models for deep fingerprints, which could support shallow machine learning methods more efficiently than traditional fingerprints or descriptors. We further performed automatic feature learning, a key essence of deep learning, to map the corresponding activation values into fragment space and derive AOT-related chemical substructures by reverse mining of the features. Our deep learning architecture for AOT is generally applicable in predicting and exploring other toxicity or property end points of chemical compounds. The two deepAOT models are freely available at http://repharma.pku.edu.cn/DLAOT/DLAOThome.php or http://www.pkumdl.cn/DLAOT/DLAOThome.php .

Statistical Analysis and Prediction of Covalent Ligand Targeted Cysteine Residues.

Targeted covalent compounds or drugs have good potency as they can bind to a specific target for a long time with low doses. Most currently known covalent ligands were discovered by chance or by modifying existing noncovalent compounds to make them covalently attached to a nearby reactive residue. Computational methods for novel covalent ligand binding prediction are highly demanded. We performed statistical analysis on protein complexes with covalent ligands attached to cysteine residues. We found that covalent modified cysteine residues have unique features compared to those not attached to covalent ligands, including lower pKa, higher exposure, and higher ligand binding affinity. SVM models were built to predict cysteine residues suitable for covalent ligand design with prediction accuracy of 0.73. Given a protein structure, our method can be used to automatically detect druggable cysteine residues for covalent ligand design, which is especially useful for identifying novel binding sites for covalent allosteric ligand design.

Cadmium modulates hematopoietic stem and progenitor cells and skews toward myelopoiesis in mice.

The heavy metal cadmium (Cd) is known to modulate immunity and cause osteoporosis. However, how Cd influences on hematopoiesis remain largely unknown. Herein, we show that wild-type C57BL/6 (B6) mice exposed to Cd for 3months had expanded bone marrow (BM) populations of long-term hematopoietic stem cells (LT-HSCs), common myeloid progenitors (CMPs) and granulocyte-macrophage progenitors (GMPs), while having reduced populations of multipotent progenitors (MPPs) and common lymphoid progenitors (CLPs). A competitive mixed BM transplantation assay indicates that BM from Cd-treated mice had impaired LT-HSC ability to differentiate into mature cells. In accordance with increased myeloid progenitors and decreased lymphoid progenitors, the BM and spleens of Cd-treated mice had more monocytes and/or neutrophils and fewer B cells and T cells. Cd impaired the ability of the non-hematopoietic system to support LT-HSCs, in that lethally irradiated Cd-treated recipients transplanted with normal BM cells had reduced LT-HSCs after the hematopoietic system was fully reconstituted. This is consistent with reduced osteoblasts, a known critical component for HSC niche, observed in Cd-treated mice. Conversely, lethally irradiated control recipients transplanted with BM cells from Cd-treated mice had normal LT-HSC reconstitution. Furthermore, both control mice and Cd-treated mice that received Alendronate, a clinical drug used for treating osteoporosis, had BM increases of LT-HSCs. Thus, the results suggest Cd increase of LT-HSCs is due to effects on HSCs and not on osteoblasts, although, Cd causes osteoblast reduction and impaired niche function for maintaining HSCs. Furthermore, Cd skews HSCs toward myelopoiesis.

Deep Learning for Drug-Induced Liver Injury.

Drug-induced liver injury (DILI) has been the single most frequent cause of safety-related drug marketing withdrawals for the past 50 years. Recently, deep learning (DL) has been successfully applied in many fields due to its exceptional and automatic learning ability. In this study, DILI prediction models were developed using DL architectures, and the best model trained on 475 drugs predicted an external validation set of 198 drugs with an accuracy of 86.9%, sensitivity of 82.5%, specificity of 92.9%, and area under the curve of 0.955, which is better than the performance of previously described DILI prediction models. Furthermore, with deep analysis, we also identified important molecular features that are related to DILI. Such DL models could improve the prediction of DILI risk in humans. The DL DILI prediction models are freely available at http://www.repharma.cn/DILIserver/DILI_home.php.