Designing antimicrobial peptides using deep learning and molecular dynamic simulations.

With the emergence of multidrug-resistant bacteria, antimicrobial peptides (AMPs) offer promising options for replacing traditional antibiotics to treat bacterial infections, but discovering and designing AMPs using traditional methods is a time-consuming and costly process. Deep learning has been applied to the de novo design of AMPs and address AMP classification with high efficiency. In this study, several natural language processing models were combined to design and identify AMPs, i.e. sequence generative adversarial nets, bidirectional encoder representations from transformers and multilayer perceptron. Then, six candidate AMPs were screened by AlphaFold2 structure prediction and molecular dynamic simulations. These peptides show low homology with known AMPs and belong to a novel class of AMPs. After initial bioactivity testing, one of the peptides, A-222, showed inhibition against gram-positive and gram-negative bacteria. The structural analysis of this novel peptide A-222 obtained by nuclear magnetic resonance confirmed the presence of an alpha-helix, which was consistent with the results predicted by AlphaFold2. We then performed a structure-activity relationship study to design a new series of peptide analogs and found that the activities of these analogs could be increased by 4-8-fold against Stenotrophomonas maltophilia WH 006 and Pseudomonas aeruginosa PAO1. Overall, deep learning shows great potential in accelerating the discovery of novel AMPs and holds promise as an important tool for developing novel AMPs.

Molecular substructure tree generative model for de novo drug design.

Deep learning shortens the cycle of the drug discovery for its success in extracting features of molecules and proteins. Generating new molecules with deep learning methods could enlarge the molecule space and obtain molecules with specific properties. However, it is also a challenging task considering that the connections between atoms are constrained by chemical rules. Aiming at generating and optimizing new valid molecules, this article proposed Molecular Substructure Tree Generative Model, in which the molecule is generated by adding substructure gradually. The proposed model is based on the Variational Auto-Encoder architecture, which uses the encoder to map molecules to the latent vector space, and then builds an autoregressive generative model as a decoder to generate new molecules from Gaussian distribution. At the same time, for the molecular optimization task, a molecular optimization model based on CycleGAN was constructed. Experiments showed that the model could generate valid and novel molecules, and the optimized model effectively improves the molecular properties.

Molecular generation strategy and optimization based on A2C reinforcement learning in de novo drug design.

MOTIVATION: In the field of pharmacochemistry, it is a time-consuming and expensive process for the new drug development. The existing drug design methods face a significant challenge in terms of generation efficiency and quality. RESULTS: In this paper, we proposed a novel molecular generation strategy and optimization based on A2C reinforcement learning. In molecular generation strategy, we adopted transformer-DNN to retain the scaffolds advantages, while accounting for the generated molecules' similarity and internal diversity by dynamic parameter adjustment, further improving the overall quality of molecule generation. In molecular optimization, we introduced heterogeneous parallel supercomputing for large-scale molecular docking based on message passing interface communication technology to rapidly obtain bioactive information, thereby enhancing the efficiency of drug design. Experiments show that our model can generate high-quality molecules with multi-objective properties at a high generation efficiency, with effectiveness and novelty close to 100%. Moreover, we used our method to assist shandong university school of pharmacy to find several candidate drugs molecules of anti-PEDV. AVAILABILITY AND IMPLEMENTATION: The datasets involved in this method and the source code are freely available to academic users at https://github.com/wq-sunshine/MomdTDSRL.git.

Large-Scale Molecular Dynamics Simulation Based on Heterogeneous Many-Core Architecture.

As the application of molecular dynamics (MD) simulations continues to evolve, the demand for accelerating large-scale simulation systems and handling of enormous simulation tasks is steadily increasing. We propose a parallel acceleration method for large-scale MD simulations based on Sunway heterogeneous many-core processors. This method integrates task scheduling, simulation calculations, and data storage, effectively tackling issues related to large-scale simulations and numerous simulation tasks. The task scheduling strategy flexibly handles tasks on various scales and enables parallel execution of multiple tasks. During the simulation calculations, we ported GROMACS to the Sunway architecture and accelerated the calculation of short-range forces through a heterogeneous processor. Our method achieves approximately 10-fold acceleration and 90% scalability when executing a single simulation task. When handling numerous simulation tasks, our method achieves parallel execution of all of the tasks with 90% scalability. By employing our method, we carried out 50 ns simulations on over 3000 distinct conotoxin structures individually within just 5 h. Additionally, we evaluated more than 200 protein-ligand complexes, and the simulation efficiency significantly exceeded that of midsized to small GPU clusters.

Research on the toxicological prognostic significance of age-related genes in endometrial cancer unveiling key factors in patient prognosis.

This study investigates the influence of aging-related genes on endometrial cancer, a prominent gynecological malignancy with rising incidence and mortality. By analyzing gene expression differences between cancerous and normal endometrial tissues, 42 aging-related genes were identified as differentially expressed. Utilizing the TCGA-UCEC sample, consensus clustering divided the samples into two molecular subgroups, Aging low and Aging high, based on their expression profiles. These subgroups showed distinct prognoses and survival rates, with the Aging high group associated with DNA repair and cell cycle pathways, and the Aging low group showing suppressed metabolic pathways and increased immune cell infiltration, suggesting a potential for better immunotherapy outcomes. Mutation analysis did not find significant differences in mutation frequencies between the groups, but a high Tumor Mutation Burden (TMB) correlated with better prognosis. A risk score model was also developed, showcasing significant prognostic power. Further analysis of the SIX1 gene revealed its overexpression in cancer cells. Drug sensitivity tests indicated that the low-risk group might respond better to chemotherapy. This research underscores the significance of aging-related genes in endometrial cancer, offering insights into their prognostic value and therapeutic potential, which could lead to personalized treatment approaches and enhanced patient management.

Endoplasmic reticulum stress promotes HBV production by enhancing use of the autophagosome/multivesicular body axis.

BACKGROUND AND AIMS: HBV infection has been reported to trigger endoplasmic reticulum (ER) stress and initiate autophagy. However, how ER stress and autophagy influence HBV production remains elusive. Here, we studied the effect of tunicamycin (TM), an N-glycosylation inhibitor and ER stress inducer, on HBV replication and secretion and examined the underlying mechanisms. APPROACH AND RESULTS: Protein disulfide isomerase (an ER marker), microtubule-associated protein 1 light chain 3 beta (an autophagosome [AP] marker), and sequestosome-1 (a typical cargo for autophagic degradation) expression were tested in liver tissues of patients with chronic HBV infection and hepatoma cell lines. The role of TM treatment in HBV production and trafficking was examined in hepatoma cell lines. TM treatment that mimics HBV infection triggered ER stress and increased AP formation, resulting in enhanced HBV replication and secretion of subviral particles (SVPs) and naked capsids. Additionally, TM reduced the number of early endosomes and HBsAg localization in this compartment, causing HBsAg/SVPs to accumulate in the ER. Thus, TM-induced AP formation serves as an alternative pathway for HBsAg/SVP trafficking. Importantly, TM inhibited AP-lysosome fusion, accompanied by enhanced AP/late endosome (LE)/multivesicular body fusion, to release HBsAg/SVPs through, or along with, exosome release. Notably, TM treatment inhibited HBsAg glycosylation, resulting in impairment of HBV virions' envelopment and secretion, but it was not critical for HBsAg/SVP trafficking in our cell systems. CONCLUSIONS: TM-induced ER stress and autophagic flux promoted HBV replication and the release of SVPs and naked capsids through the AP-LE/MVB axis.

Factors related to dietary quality among older stroke high-risk population in Tianjin community, China: a multicenter study.

BACKGROUND: Stroke is a common and frequently-occurring disease in older people. It has the characteristics of high morbidity, high mortality, high recurrence rate and high disability rate. Most stroke risk studies are based on pathophysiology, however psychosocial factors such as diet quality are often understudied. The aim of this study was to assess stroke risk in urban community residents in Tianjin and investigate the factors that affect the dietary quality of older stroke high-risk populations. METHODS: Using a cross-sectional, multicenter study, recruit people aged 60 to 80 in Tianjin. Dietary intake data were obtained through a validated food frequency questionnaire, which were used to calculate Alternate Healthy Eating Index-2010 (AHEI-2010) and to analyze its association with sociodemographic characteristics, stroke risk factors and health marker variables. RESULTS: A total of 1068 participants from 4 community health service centers in Tianjin were recruited, including 300 low-risk individuals and 768 high-risk individuals. Compared with the low-risk group (62.75 ± 3.59), the AHEI-2010 mean score of the high-risk group (56.83 ± 6.54) was significantly lower. The top three most common risk factors among participants were dyslipidemia (80.3%), hypertension (60.6%), and physical inactivity (58.2%). Multiple logistic regression showed that diet quality was independently and significantly associated with stroke risk (OR = 0.765; 95%CI: 0.690-0.848, p < 0.001). CONCLUSION: The diet quality of high-risk stroke population in Tianjin is far from ideal. At the same time, public health knowledge needs to be disseminated and educated, especially among those at high risk of cerebrovascular disease, with a focus on improving psychosocial factors such as diet quality.

Attractor radius for fractional Lorenz systems and their application to the quantification of predictability limits.

Quantifying the predictability limits of chaotic systems and their forecast models has attracted much interest among scientists. The attractor radius (AR) and the global attractor radius (GAR), as intrinsic properties of a chaotic system, were introduced in the most recent work (Li et al. 2018). It has been shown that both the AR and GAR provide more accurate, objective metrics to access the global and local predictability limits of forecast models compared with the traditional error saturation or the asymptotic value. In this work, we consider the AR and GAR of fractional Lorenz systems, introduced in Grigorenko and Grigorenko [Phys. Rev. Lett. 91, 034101 (2003)] using the Caputo fractional derivatives and their application to the quantification of the predictability limits. A striking finding is that a fractional Lorenz system with smaller Σ, which is a sum of the orders of all involved equal derivatives, has smaller attractor radius and shorter predictability limits. In addition, we present a new numerical algorithm for the fractional Lorenz system, which is the generalized version of the standard fourth-order Runge-Kutta scheme.

An odorant receptor tuned to an attractive plant volatile vanillin in Spodoptera litura.

The insect olfaction plays crucial roles in many important behaviors, in which ORs are key determinants for signal transduction and the olfactory specificity. Spodoptera litura is a typical polyphagous pest, possessing a large repertoire of ORs tuning to broad range of plant odorants. However, the specific functions of those ORs remain mostly unknown. In this study, we functionally characterized one S. litura OR (OR51) that was highly expressed in the adult antennae. First, by using Xenopus oocyte expression and two-electrode voltage clamp recording system (XOE-TEVC), OR51 was found to be strongly and specifically responsive to vanillin (a volatile of S. litura host plants) among 77 tested odorants. Second, electroantennogram (EAG) and Y-tube behavioral experiment showed that vanillin elicited significant EAG response and attraction behavior especially of female adults. This female attraction was further confirmed by the oviposition experiment, in which the soybean plants treated with vanillin were significantly preferred by females for egg-laying. Third, 3D structural modelling and molecular docking were conducted to explore the interaction between OR51 and vanillin, which showed a high affinity (-4.46 kcal/mol) and three residues (Gln163, Phe164 and Ala305) forming hydrogen bonds with vanillin, supporting the specific binding of OR51 to vanillin. In addition, OR51 and its homologs from other seven noctuid species shared high amino acid identities (78-97%) and the same three hydrogen bond forming residues, suggesting a conserved function of the OR in these insects. Taken together, our study provides some new insights into the olfactory mechanisms of host plant finding and suggests potential applications of vanillin in S. litura control.

GraphGPT: A Graph Enhanced Generative Pretrained Transformer for Conditioned Molecular Generation.

Condition-based molecular generation can generate a large number of molecules with particular properties, expanding the virtual drug screening library, and accelerating the process of drug discovery. In this study, we combined a molecular graph structure and sequential representations using a generative pretrained transformer (GPT) architecture for generating molecules conditionally. The incorporation of graph structure information facilitated a better comprehension of molecular topological features, and the augmentation of a sequential contextual understanding of GPT architecture facilitated molecular generation. The experiments indicate that our model efficiently produces molecules with the desired properties, with valid and unique metrics that are close to 100%. Faced with the typical task of generating molecules based on a scaffold in drug discovery, our model is able to preserve scaffold information and generate molecules with low similarity and specified properties.

Copper-Catalyzed Enantioselective Difluoromethylation-Alkynylation of Olefins by Solving the Dilemma between Acidities and Reduction Potentials of Difluoromethylating Agents.

Molecules containing a difluoromethyl group or a propargylic stereocenter are widely used in pharmaceuticals and agrochemicals, and 1,2-functionalization of olefins is an important method for introducing the two groups into molecules simultaneously. The construction of the propargylic stereocenter with terminal alkynes usually requires bases. However, difluoromethylating agents with high reduction potentials often decompose in the presence of bases because of their acidities, and those with low reduction potentials are stable but difficult to undergo the desired single electron transfer (SET) reduction. Using the linear relationship between reduction potential differences (ΔE) and Hammett substituent constants (σ) of difluoromethyl aryl sulfones, we solved the dilemma between acidities and reduction potentials of difluoromethylating agents. Herein, we report the first enantioselective difluoromethylation-alkynylation of olefins with difluoromethyl 4-chlorophenyl sulfone with high enantioselectivity (>90 % ee). We also extended this asymmetric fluoroalkylation-alkynylation reaction with other fluoroalkyl sulfones, which enabled efficient installation of trifluoromethyl, difluoroalkyl, difluorobenzyl, (benzenesulfonyl)-difluoromethyl and monofluoromethyl groups into products.

Regioselective Aromatic Perfluoro-tert-butylation Using Perfluoro-tert-butyl Phenyl Sulfone and Arynes.

The selective introduction of perfluoro-tert-butyl group (PFtB, the bulkier analogue of CF3 group) into arenes has long been sought after but remains a formidable task. We herein report the first general synthetic protocol to realize aromatic perfluoro-tert-butylation. The key to the success is the identification of PFtB phenyl sulfone as a new source of PFtB anion, which reacts with arynes in a highly regioselective manner to afford perfluoro-tert-butylated arenes in high yields. The application of the method is demonstrated by the preparation of sensitive 19F-labeled NMR probes with an extraordinary resolving ability.

MHBSt167 induced autophagy promote cell proliferation and EMT by activating the immune response in L02 cells.

BACKGROUND: Hepatitis B virus can induce hepatocellular carcinoma (HCC) by inducing a host immune response against infected hepatocytes. C-terminally truncated middle surface protein (MHBSt) has been reported to contribute to HCC through transcriptional activation in epidemiology studies, while the underlying mechanism of MHBSt-induced HCC is unknown. METHODS: In this study, a premature stop at codon 167 in MHBS (MHBSt167) was investigated into eukaryotic expression plasmid pcDNA3.1(-). MHBSt167 expressed plasmid was transfected into the L02 cell line, cell proliferation was analyzed by CCK-8 and high-content screening assays, the cell cycle was analyzed by flow cytometry, and epithelial-to-mesenchymal transition and autophagy were analyzed by immunoblotting and immunofluorescence. NF-κB activation and the MHBSt167-induced immune response were analyzed by immunoblotting and immunofluorescence. IFN-α, IFN-ß and IL-1α expression were analyzed by qPCR. Autophagy inhibitors were used to analyze the relationship between the immune response and autophagy. RESULTS: The results showed that MHBSt167 promoted L02 cell proliferation, accelerated cell cycle progression from the S to G2 phase and promoted epithelial-to-mesenchymal transition through ER-stress, leading to autophagy and NF-κB activation and increased immune-related factor expression. The MHBSt167-induced acceleration of cell proliferation and the cell cycle was abolished by autophagy or NF-κB inhibitors. CONCLUSION: In summary, MHBSt167 could promote cell proliferation, accelerate cell cycle progression, induce EMT and activate autophagy through ER-stress to induce the host immune response, supporting a potential role of MHBSt167 in contributing to carcinogenesis.

Identification of the miR-423-3p/VLDLR Regulatory Network for Glioma Using Transcriptome Analysis.

As the most prevalent primary CNS tumor, glioma is characterized by high mortality and morbidity. This research aims to investigate glioma-associated microRNAs (miRNAs) and their target mRNAs, as well as to explore their biological functions in gliomas. The Gene Expression Omnibus (GEO) database was applied to acquire the GSE112264 miRNA microarray dataset and the GSE15824 mRNA dataset. We selected samples from the GSE112264 dataset and the GSE15824 to identify differently expressed miRNAs (DE-miRNAs) as well as differentially expressed mRNAs (DEGs), respectively. Next, the intersections of mRNA and target mRNAs of miRNA were selected, and we constructed miRNA-mRNA regulation networks. These DEGs were selected for Gene Oncology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses by conducting the package clusterProfiler. After conducting Cytoscape software, a protein-protein interaction (PPI) network was created. Next, survival analysis of the miR-423-3p was confirmed by conducting TCGA database. Subsequently, Quantitative real-time PCR (qRT-PCR) was conducted to verify miR-423-3p's expression. Finally, miR-423-3p's biological functions of in effecting the cell proliferative, migratory, and invasive capabilities of glioma were investigated by performing Cell Counting Kit-8 (CCK-8) and Transwell assays. Our analysis elucidated a novel miRNA-mRNA regulatory network related to glioma carcinogenesis, which may be considered as future therapeutic biomarkers for glioma.

Enhancing Protein Function Prediction Performance by Utilizing AlphaFold-Predicted Protein Structures.

The structure of a protein is of great importance in determining its functionality, and this characteristic can be leveraged to train data-driven prediction models. However, the limited number of available protein structures severely limits the performance of these models. AlphaFold2 and its open-source data set of predicted protein structures have provided a promising solution to this problem, and these predicted structures are expected to benefit the model performance by increasing the number of training samples. In this work, we constructed a new data set that acted as a benchmark and implemented a state-of-the-art structure-based approach for determining whether the performance of the function prediction model can be improved by putting additional AlphaFold-predicted structures into the training set and further compared the performance differences between two models separately trained with real structures only and AlphaFold-predicted structures only. Experimental results indicated that structure-based protein function prediction models could benefit from virtual training data consisting of AlphaFold-predicted structures. First, model performances were improved in all three categories of Gene Ontology terms (GO terms) after adding predicted structures as training samples. Second, the model trained only on AlphaFold-predicted virtual samples achieved comparable performances to the model based on experimentally solved real structures, suggesting that predicted structures were almost equally effective in predicting protein functionality.

Functional differentiation of three pheromone binding proteins in Orthaga achatina using mixed-type sex pheromones.

Pheromone-binding proteins (PBPs) play important roles in perception of insect sex pheromones, functioning to recognize and transport pheromone components onto the olfactory receptors of the odorant sensing neurons. Orthaga achatina, a serious pest of camphor trees, uses a mixture of three Type I (Z11-16:OAc, Z11-16:OH and Z11-16:Ald) and one Type II (Z3,Z6,Z9,Z12,Z15-23:H) sex pheromone components in its sex communication, in which Z11-16:OAc is the major component and others are minor components. In this study, we for the first time demonstrated that the three PBPs differentiated in recognition among pheromone components in a moth using mixed-type sex pheromones. First, tissue expression study showed that all three PBPs of O. achatina were expressed only in antennae and highly male-biased, suggesting their involvement in perception of the sex pheromones. Second, the three PBPs were expressed in Escherichia coli and the binding affinities of PBPs to four sex pheromone components and some pheromone analogs were determined by the fluorescence competition binding assays. The results showed that OachPBP1 bound all four sex pheromone components with high binding affinity, while OachPBP2 had high or moderate binding affinity only to three Type I components, and OachPBP3 had high binding affinity only to three minor pheromone components. Furthermore, key amino acid residues that bind to sex pheromone components were identified in three PBPs by 3-D structure modeling and ligand molecular docking, predicting the interactions between PBPs and pheromone components. Our study provides a fundamental insight into the olfactory mechanism in moths that use mixed-type sex pheromones.

Few-Shot Object Detection Using Multimodal Sensor Systems of Unmanned Surface Vehicles.

The object detection algorithm is a key component for the autonomous operation of unmanned surface vehicles (USVs). However, owing to complex marine conditions, it is difficult to obtain large-scale, fully labeled surface object datasets. Shipborne sensors are often susceptible to external interference and have unsatisfying performance, compromising the results of traditional object detection tasks. In this paper, a few-shot surface object detection method is proposed based on multimodal sensor systems for USVs. The multi-modal sensors were used for three-dimensional object detection, and the ability of USVs to detect moving objects was enhanced, realizing metric learning-based few-shot object detection for USVs. Compared with conventional methods, the proposed method enhanced the classification results of few-shot tasks. The proposed approach achieves relatively better performance in three sampled sets of well-known datasets, i.e., 2%, 10%, 5% on average precision (AP) and 28%, 24%, 24% on average orientation similarity (AOS). Therefore, this study can be potentially used for various applications where the number of labeled data is not enough to acquire a compromising result.

Immunostimulatory siRNA with a uridine bulge leads to potent inhibition of HBV and activation of innate immunity.

BACKGROUND: Hepatitis B virus (HBV) infection is difficult to cure. HBV-specific immune tolerance plays a key role in HBV persistence, and enhancing cellular and humoral immunity will improve the control of HBV infection. The purpose of the study was to explore the anti-HBV and immunostimulatory effects of msiRNAs that introduce unpaired uridine bulges in the passenger strand. METHODS: msiRNAs targeting the HBV S and X genes were designed and named msiHBs and msiHBx, respectively. HepG2 cells were cotransfected with siRNA or msiRNA and the HBV replication-competent plasmid pHY106-wta or pHY106-X15. HepG2.215 cells were transfected with siRNA or msiRNA. The levels of HBsAg, HBeAg, and the cytokines TNF-α, IFN-α, IFN-ß, IL-1α, and IL-6 in the culture supernatant was detected by ELISA. The levels of intracellular HBV RNA, nuclear HBV replication intermediates, and HBV DNA in the supernatant were measured by quantitative RT-PCR and PCR. The levels of HBV replication intermediates were detected by Southern blotting. Peripheral blood mononuclear cells were transfected with siRNA or msiRNA, and the levels of secreted cytokines IFN-α and IFN-ß were detected by ELISA. The bioactivity of type I interferons in the supernatants was detected by the virus protection assay. RESULTS: msiHBx treatment led to a significant decrease in HBsAg (to a negative level) and HBV DNA (95.5%) in the supernatant and intrahepatocellular HBV replication intermediates (89.8%) in HepG2 cells with transient HBV replication and in HepG2.2.15 cells. There was no significant difference between msiHBx and siHBx in terms of the reduction in HBV proteins and HBV replication (P > 0.05). Compared with siHBx, msiHBx treatment of HepG2 cells transfected with the HBV replication-competent plasmid led to a significant increase in the levels of the antiviral cytokines TNF-α (3.3-fold), IFN-α (1.4-fold), and IFN-ß (2.5-fold) (P < 0.01), without upregulation of the proinflammatory cytokines IL-1α and IL-6. The virus protection assay results showed msiHBx-mediated type I interferons effectively protected L929 cells against ECMV infection. CONCLUSIONS: msiHBx could effectively inhibit HBV expression and replication and induce an antiviral innate immune response without proinflammatory activation. The dual RNAi and immunostimulatory activity of msiRNAs may play an important role in the control of HBV infection.

Sensor-Fusion for Smartphone Location Tracking Using Hybrid Multimodal Deep Neural Networks.

Many engineered approaches have been proposed over the years for solving the hard problem of performing indoor localization using smartphone sensors. However, specialising these solutions for difficult edge cases remains challenging. Here we propose an end-to-end hybrid multimodal deep neural network localization system, MM-Loc, relying on zero hand-engineered features, but learning automatically from data instead. This is achieved by using modality-specific neural networks to extract preliminary features from each sensing modality, which are then combined by cross-modality neural structures. We show that our choice of modality-specific neural architectures can estimate the location independently. But for better accuracy, a multimodal neural network that fuses the features of early modality-specific representations is a better proposition. Our proposed MM-Loc system is tested on cross-modality samples characterised by different sampling rate and data representation (inertial sensors, magnetic and WiFi signals), outperforming traditional approaches for location estimation. MM-Loc elegantly trains directly from data unlike conventional indoor positioning systems, which rely on human intuition.

SAG-DTA: Prediction of Drug-Target Affinity Using Self-Attention Graph Network.

The prediction of drug-target affinity (DTA) is a crucial step for drug screening and discovery. In this study, a new graph-based prediction model named SAG-DTA (self-attention graph drug-target affinity) was implemented. Unlike previous graph-based methods, the proposed model utilized self-attention mechanisms on the drug molecular graph to obtain effective representations of drugs for DTA prediction. Features of each atom node in the molecular graph were weighted using an attention score before being aggregated as molecule representation. Various self-attention scoring methods were compared in this study. In addition, two pooing architectures, namely, global and hierarchical architectures, were presented and evaluated on benchmark datasets. Results of comparative experiments on both regression and binary classification tasks showed that SAG-DTA was superior to previous sequence-based or other graph-based methods and exhibited good generalization ability.