Protein Sequence Research Articles

The role of silent mutations in KRAS-mutant tumors.

Silent mutations within the RASgene have garnered increasing attention for their potential roles in tumorigenesis and therapeutic strategies. Kirsten-RAS (KRAS)mutations, predominantly oncogenic, are pivotal drivers in various cancers. While extensive research has elucidated the molecular mechanisms and biological consequences of active KRASmutations, the functional significance of silent mutations remains relatively understudied. This review synthesizes current knowledge on KRASsilent mutations, highlighting their impact on cancer development. Silent mutations, which do not alter protein sequences but can affect RNA stability and translational efficiency, pose intriguing questions regarding their contribution to tumor biology. Understanding these mutations is crucial for comprehensively unraveling KRAS-driven oncogenesis and exploring novel therapeutic avenues. Moreover, investigations into the clinical implications of silent mutations in KRAS-mutant tumors suggest potential diagnostic and therapeutic strategies. Despite being in early stages, research on KRASsilent mutations holds promise for uncovering novel insights that could inform personalized cancer treatments. In conclusion, this review underscores the evolving landscape of KRASsilent mutations, advocating for further exploration to bridge fundamental biology with clinical applications in oncology.

Designing host-associated microbiomes using the consumer/resource model.

Generative models are extremely popular in modern biology. They have been used to model the variation of protein sequences, entire genomes, and RNA sequencing profiles. Importantly, generative models have been used to extrapolate and interpolate to unobserved regimes of data to design biological systems with desired properties. For example, there has been a boom in machine-learning models aiding in the design of proteins with user-specified structures or functions. Host-associated microbiomes play important roles in animal health and disease, as well as the productivity and environmental footprint of livestock species. However, there are no generative models of host-associated microbiomes. One chief reason is that off-the-shelf machine-learning models are data hungry, and microbiome studies usually deal with large variability and small sample sizes. Moreover, microbiome compositions are heavily context dependent, with characteristics of the host and the abiotic environment leading to distinct patterns in host-microbiome associations. Consequently, off-the-shelf generative modeling has not been successfully applied to microbiomes.To address these challenges, we develop a generative model for host-associated microbiomes derived from the consumer/resource (C/R) framework. This derivation allows us to fit the model to readily available cross-sectional microbiome profile data. Using data from three animal hosts, we show that this mechanistic generative model has several salient features: the model identifies a latent space that represents variables that determine the growth and, therefore, relative abundances of microbial species. Probabilistic modeling of variation in this latent space allows us to generate realistic in silico microbial communities. The model can assign probabilities to microbiomes, thereby allowing us to discriminate between dissimilar ecosystems. Importantly, the model predictively captures host-associated microbiomes and the corresponding hosts' phenotypes, enabling the design of microbial communities associated with user-specified host characteristics.

Redesigning error control in cross-linking mass spectrometry enables more robust and sensitive protein-protein interaction studies.

Cross-linking mass spectrometry (XL-MS) allows characterizing protein-protein interactions (PPIs) in native biological systems by capturing cross-links between different proteins (inter-links). However, inter-link identification remains challenging, requiring dedicated data filtering schemes and thorough error control. Here, we benchmark existing data filtering schemes combined with error rate estimation strategies utilizing concatenated target-decoy protein sequence databases. These workflows show shortcomings either in sensitivity (many false negatives) or specificity (many false positives). To ameliorate the limited sensitivity without compromising specificity, we develop an alternative target-decoy search strategy using fused target-decoy databases. Furthermore, we devise a different data filtering scheme that takes the inter-link context of the XL-MS dataset into account. Combining both approaches maintains low error rates and minimizes false negatives, as we show by mathematical simulations, analysis of experimental ground-truth data, and application to various biological datasets. In human cells, inter-link identifications increase by 75% and we confirm their structural accuracy through proteome-wide comparisons to AlphaFold2-derived models. Taken together, target-decoy fusion and context-sensitive data filtering deepen and fine-tune XL-MS-based interactomics.

ProHap enables human proteomic database generation accounting for population diversity.

Amid the advances in genomics, the availability of large reference panels of human haplotypes is key to account for human diversity within and across populations. However, mass spectrometry-based proteomics does not benefit from this information. To address this gap, we introduce ProHap, a Python-based tool that constructs protein sequence databases from phased genotypes of reference panels. ProHap enables researchers to account for haplotype diversity in proteomic searches.

TopDIA: A Software Tool for Top-Down Data-Independent Acquisition Proteomics.

Top-down mass spectrometry is widely used for proteoform identification, characterization, and quantification owing to its ability to analyze intact proteoforms. In the past decade, top-down proteomics has been dominated by top-down data-dependent acquisition mass spectrometry (TD-DDA-MS), and top-down data-independent acquisition mass spectrometry (TD-DIA-MS) has not been well studied. While TD-DIA-MS produces complex multiplexed tandem mass spectrometry (MS/MS) spectra, which are challenging to confidently identify, it selects more precursor ions for MS/MS analysis and has the potential to increase proteoform identifications compared with TD-DDA-MS. Here we present TopDIA, the first software tool for proteoform identification by TD-DIA-MS. It generates demultiplexed pseudo MS/MS spectra from TD-DIA-MS data and then searches the pseudo MS/MS spectra against a protein sequence database for proteoform identification. We compared the performance of TD-DDA-MS and TD-DIA-MS using Escherichia coli K-12 MG1655 cells and demonstrated that TD-DIA-MS with TopDIA increased proteoform and protein identifications compared with TD-DDA-MS.

Genome-wide identification, gene cloning, subcellular location and expression analysis of the OPR gene family under salt stress in sweetpotato

BackgroundThe 12-oxo-phytodienoic acid reductase (OPR) enzyme is crucial for the synthesis of jasmonates (JAs), and is involved in the plant stress response. However, the OPR gene family in sweetpotato, an important horticultural crop, remains unidentified.ResultsIn this study, we employed bioinformatics techniques to identify nine IbOPR genes. Phylogenetic analysis revealed that these genes could be divided into Group I and Group II. Synteny analysis indicated that IbOPR evolution was driven by tandem duplication, whole-genome duplication (WGD), and segmental duplication events. The promoter sequences of IbOPRs were found to be associated with stress and hormonal responses. Additionally, we successfully cloned four IbOPRs from "Haida HD7791" and "Haida HD7798" using homologous cloning technology. These sequences were 1203 bp, 1200 bp, 1134 bp, and 1137 bp in length and encoded 400, 399, 377, and 378 amino acids, respectively. The protein sequence similarity between the salt-tolerant variety "Haida HD7791" and the salt-sensitive variety "Haida HD7798" was determined to be 96.75% for IbOPR2, 99.75% for IbOPR3, 92.06% for IbOPR6, and 98.68% for IbOPR7. Phylogenetic analysis categorized IbOPR2 and IbOPR3 proteins into Group II, while IbOPR6 and IbOPR7 proteins belonged to Group I. Subcellular localization experiments showed IbOPR2 protein present in the peroxisome, while IbOPR3, IbOPR6, and IbOPR7 proteins were found in the cytoplasm and nucleus. Salt stress induction experiments demonstrated that IbOPR2, IbOPR3, and IbOPR7 were significantly upregulated only in 'Haida HD7791' after 6 h. In contrast, IbOPR6 was induced in 'Haida HD7798' at 6 h but inhibited in 'Haida HD7791' at later time points (12, 24, 48, and 72 h), highlighting functional differences in salt stress responses.ConclusionsOur findings suggest that IbOPR2 may play a crucial role in sweetpotato's response to salt stress by participating in JAs synthesis. These results provide a foundation for future functional analyses of OPR genes in sweetpotato.

Effects of energetic compounds on soil microbial communities and functional genes at a typical ammunition demolition site.

High concentrations of energetic compounds such as 2,4,6-trinitrotoluene (TNT), hexahydro-1,3,5-trinitro-1,3,5-triazine (RDX), and octahydro-1,3,5,7-tetranitro-1,3,5,7-tetrazocine (HMX) in military-contaminated sites pose a serious threat to human health and ecosystems. Better understanding about their effects on microbial diversity and functional genes in soil of ammunition demolition sites is required. In this study, the information of soil microbial community composition was obtained by macrogenome sequencing, and the impacts of energetic compounds on microbial community structure at the level of functional genes and enzymes based on Nr (Non-Redundant Protein Sequence Database), KEGG (Kyoto Encyclopedia of Genes and Genomes), CAZy (Carbohydrate-Active En-zymes Database) and other databases were discussed. The results showed that soil microbial diversity and functional gene abundance decreased significantly with the increase of the concentrations of energetic compounds. Conversely, the relative abundance of Proteobacteria increased significantly, reaching over 80% in the heavily TNT-contaminated area near explosive-waste water pool. Furthermore, functional gene analysis indicated that Proteobacteria had an advantage in degrading energetic compounds, and thus had the potential to improve the soil quality at ammunition demolition sites. This study provides a scientific basis for the future remediation and management of contaminated soils at ammunition demolition sites, as well as for the selection of efficient degraders of energetic compounds.

Genetic Analysis of the Peach SnRK1β3 Subunit and Its Function in Transgenic Tomato Plants

Background/Objectives: The sucrose non-fermentation-related kinase 1 (SnRK1) protein complex in plants plays an important role in energy metabolism, anabolism, growth, and stress resistance. SnRK1 is a heterotrimeric complex. The SnRK1 complex is mainly composed of α, β, βγ, and γ subunits. Studies on plant SnRK1 have primarily focused on the functional α subunit, with the β regulatory subunit remaining relatively unexplored. The present study aimed to elucidate the evolutionary relationship, structural prediction, and interaction with the core α subunit of peach SnRK1β3 (PpSnRK1) subunit. Methods: Bioinformatics analysis of PpSnRK1 was performed through software and website. We produced transgenic tomato plants overexpressing PpSnRK1 (OEPpSnRK1). Transcriptome analysis was performed on OEPpSnRK1 tomatoes. We mainly tested the growth index and drought resistance of transgenic tomato plants. Results: The results showed that PpSnRK1 has a 354 bp encoded protein sequence (cds), which is mainly located in the nucleus and cell membrane. Phylogenetic tree analysis showed that PpSnRK1β3 has similar domains to other woody plants. Transcriptome analysis of OEPpSnRK1β3 showed that PpSnRK1β3 is widely involved in biosynthetic and metabolic processes. Functional analyses of these transgenic plants revealed prolonged growth periods, enhanced growth potential, improved photosynthetic activity, and superior drought stress tolerance. Conclusions: The study findings provide insight into the function of the PpSnRK1 subunit and its potential role in regulating plant growth and drought responses. This comprehensive analysis of PpSnRK1 will contribute to further enhancing our understanding of the plant SnRK1 protein complex.

Dominance of recombinant DWV genomes with changing viral landscapes as revealed in national US honey bee and varroa mite survey

Honey bees are essential pollinators for global agriculture. The viromes of US commercial apiaries and their ectoparasitic mites are poorly characterized at a strain level and there is a need to integrate genomics into pathogen surveillance. We sequenced RNA viromes from 383 adult bees and 173 mites pooled samples from 11 major US beekeeping hubs in 2021, assembling 45 complete and 1702 partial genomes. Protein sequence similarity networks and recombinant genome identification revealed a new viral landscape. Sinaivirus (n = 312), Iflavirus sacbroodi (n = 280), and Iflavirus aladeformis (DWV, n = 135) genomes were common. Recombinant DWV genomes with high nucleotide identity were widespread, and DWV type A master variants were rare, with an indication that RT-PCR surveillance may over-detect type A due to the prevalence of recombinant DWV genomes. Future work should use genomic strategies to avoid misidentification of common honey bee virus genomes and their impact on colony health.

The identification of a SARs-CoV2 S2 protein-derived peptide with super-antigen-like stimulatory properties on T-cells.

Severe COVID-19 can trigger a cytokine storm, leading to acute respiratory distress syndrome (ARDS) with similarities to superantigen-induced toxic shock syndrome. An outstanding question is whether SARS-CoV-2 protein sequences can directly induce inflammatory responses. In this study, we identify a region in the SARS-CoV-2 S2 spike protein with sequence homology to bacterial super-antigens (termed P3). Computational modeling predicts P3 binding to sites on MHC class I/II and the TCR that partially overlap with sites for the binding of staphylococcal enterotoxins B and H. Like SEB and SEH peptides, P3 stimulated 25-40% of human CD4+ and CD8+ T cells, increasing IFN-γ and granzyme B production. viSNE and SPADE profiling identified overlapping and distinct IFN-γ and GZMB subsets. The super-antigenic properties of P3 were further evident by its selective expansion of T cells expressing specific TCR Vα and Vβ chain repertoires. In vivo experiments in mice revealed that the administration of P3 led to a significant upregulation of proinflammatory cytokines IL-1β, IL-6, and TNF-α. While the clinical significance of P3 in COVID-19 remains unclear, its homology to other mammalian proteins suggests a potential role for this peptide family in human inflammation and autoimmunity.

Borrelia PeptideAtlas: A proteome resource of common Borrelia burgdorferi isolates for Lyme research

Lyme disease is caused by an infection with the spirochete Borrelia burgdorferi, and is the most common vector-borne disease in North America. B. burgdorferi isolates harbor extensive genomic and proteomic variability and further comparison of isolates is key to understanding the infectivity of the spirochetes and biological impacts of identified sequence variants. Here, we applied both transcriptome analysis and mass spectrometry-based proteomics to assemble peptide datasets of B. burgdorferi laboratory isolates B31, MM1, and the infective isolate B31-5A4, to provide a publicly available Borrelia PeptideAtlas. Included are total proteome, secretome, and membrane proteome identifications of the individual isolates. Proteomic data collected from 35 different experiment datasets, totaling 386 mass spectrometry runs, have identified 81,967 distinct peptides, which map to 1,113 proteins. The Borrelia PeptideAtlas covers 86% of the total B31 proteome of 1,291 protein sequences. The Borrelia PeptideAtlas is an extensible comprehensive peptide repository with proteomic information from B. burgdorferi isolates useful for Lyme disease research.

Assessment of Punicalagin activity Against Influenza a Virus Proteins via In-silico Approaches

Background: To investigate the anti-influenza mechanism of punicalagin and its inhibitory effects on influenza virus proteins, we conducted molecular docking studies targeting 11 viral proteins. Additionally, molecular dynamics simulations were performed for the protein with the lowest free energy and the minimum concentration required for binding in a simulated environment. Methods: The molecular structure and data of punicalagin were obtained from the PubChem database and converted into a PDB file. FASTA sequences of viral proteins were retrieved from UniProt, and their PDB structures were predicted using the I-TASSER server. Molecular docking was performed using AutoDock 4.2 software, while molecular dynamics simulations were conducted with Gromacs 2022 software. Results: The PB1-F2 protein exhibited the best inhibitory performance, with a binding energy of -5.76 kcal/mol and the lowest inhibition constant (Ki). Docking of punicalagin to the PB1-F2 protein led to a significant decrease in the average total energy (TE), radius of gyration (Rg), and root mean square deviation (RMSD), as well as alterations in the secondary structure of the protein (P < 0.001). The most prominent secondary structural change was a reduction in coil structures and an increase in turn structures. Conclusions: Punicalagin displayed a strong binding affinity for the PB1-F2 protein compared to other influenza viral proteins. Considering the role of PB1-F2 in exacerbating inflammation caused by influenza, punicalagin may mitigate inflammation in patients by modulating the activity of this protein.

Genome-wide analysis and expression profiling of the polyamine oxidase gene family in Solanum tuberosum L.

BackgroundPolyamine oxidase (PAO) is a crucial enzyme involved in the breakdown of polyamines (PAs) in plants. It not only regulates the levels of PAs, but also plays a role in the oxidative decomposition of PAs and the release of stress-related signals, contributing to the plant’s response and resistance to various adversities. While there have been numerous studies on the response of PAO to stress in other crops, there is a lack of research on this topic in potatoes, a major food crop.ResultsIn this study, we aimed to explore the biological function of the StPAO gene in potato growth and development, as well as its expression patterns under stress. Using bioinformatics methods, we identified 14 StPAO genes in the potato genome. Protein sequence comparisons revealed a high similarity between the PAO proteins of potato and Arabidopsis. Chromosomal mapping and gene structure analysis showed that the StPAO genes were not evenly distributed on the chromosome and all contained an amino-oxidase domain. Furthermore, analysis of the promoters of these genes revealed the presence of abiotic and stress-related cis-acting elements, indicating their potential role in responding to different stresses. To investigate the expression patterns of these genes under stress, we used qRT-PCR to study their response to high temperature, drought, and ABA stress. Our results showed that StPAO6 and StPAO10 were significantly up-regulated under high temperature stress, indicating that they were involved in the process of potato resistance to high temperatures. Similarly, StPAO1, StPAO3, and StPAO4 were significantly up-regulated under drought stress, indicating their potential role in potatoes’ responses to drought. After ABA treatment, the expression levels of StPAO4, StPAO5, StPAO7, and StPAO14 were significantly up-regulated, suggesting their involvement in chemical defense mechanisms. Interestingly, the expression of StPAO11–13 was inhibited by all three stresses.ConclusionsIn conclusion, our study highlights the multifunctional nature of the StPAO gene family in potatoes, which plays a crucial role in coping with various stresses. This research deepens our understanding of the potato StPAO gene family and provides a reference for future studies on its function. It also serves as a theoretical basis for breeding stress-resistant potato varieties in the future.

Unveiling DNA Sequences: A Comparison of Machine Learning and Deep Learning Techniques for Prediction

<div align="center"><table width="590" border="1" cellspacing="0" cellpadding="0"><tbody><tr><td valign="top" width="387"><p>DNA is the biological macromolecule unit that carries the information of all protein, amino acid sequences. With the help of this protein sequence, we explore the mutated gene and disease-causing mutated genomic pattern. Currently, the progression of genomic innovation is the source of DNA arrangement information developing at a dangerous rate—external factors have stimulated the volume of research into DNA genomes. Initially, the development process of DNA sequencing is accomplished with the support of the Database, data structures, and sequence similarity. The method is capable of extracting a particular property in DNA. We employ the deep learning algorithm to pull out protein sequences' features. The DNA sequence is classified based on the in-build protein structures extracted into the Fasta file. Therefore, the DNA sequence of E.coli with 106 data sets and 57 nucleotides is tested experimentally. Finally, we compared the results with the existing decision tree algorithm, KNN- Classification, random forest, and neural networks. The deep learning algorithm yields higher efficiency of 98% compared to other machine learning algorithms. This highlights the potential of deep learning in genomics research and its ability to yield superior results in classifying DNA sequences.</p></td></tr></tbody></table></div>

Label Deconvolution for Node Representation Learning on Large-Scale Attributed Graphs Against Learning Bias.

Node representation learning on attributed graphs-whose nodes are associated with rich attributes (e.g., texts and protein sequences)-plays a crucial role in many important downstream tasks. To encode the attributes and graph structures simultaneously, recent studies integrate pre-trained models with graph neural networks (GNNs), where pre-trained models serve as node encoders (NEs) to encode the attributes. As jointly training large NEs and GNNs on large-scale graphs suffers from severe scalability issues, many methods propose to train NEs and GNNs separately. Consequently, they do not take feature convolutions in GNNs into consideration in the training phase of NEs, leading to a significant learning bias relative to the joint training. To address this challenge, we propose an efficient label regularization technique, namely Label Deconvolution (LD), to alleviate the learning bias by a novel and highly scalable approximation to the inverse mapping of GNNs. The inverse mapping leads to an objective function that is equivalent to that by the joint training, while it can effectively incorporate GNNs in the training phase of NEs against the learning bias. More importantly, we show that LD converges to the optimal objective function values by the joint training under mild assumptions. Experiments demonstrate LD significantly outperforms state-of-the-art methods on Open Graph Benchmark datasets.

GACT-PPIS: Prediction of protein-protein interaction sites based on graph structure and transformer network

The prediction of protein-protein interaction sites (PPIS) is currently crucial for regulating many biological activities in cells and developing drugs for various diseases. Deep learning-based methods have been proposed for predicting PPIS, significantly reducing the manpower and time costs associated with traditional experimental methods such as yeast two-hybrid, mass spectrometry, and affinity purification. However, the predictive accuracy of these deep learning methods has not yet reached the expected level. Therefore, we introduce a model called GACT-PPIS.The design of the GACT-PPIS algorithm aims to utilize combined information from protein sequences and structures as input to predict protein-protein interaction sites. The core of GACT-PPIS utilizes an Enhanced Graph Attention Network (EGAT) with initial residual and identity mappings, along with a deep Transformer network as the basic units, supplemented by Graph Convolutional Networks (GCN), effectively aggregating information from neighboring nodes for each node. After multiple network layers, the information of the entire protein is also fused into the nodes, and the Transformer network further enhances the model's performance.Experimental results show that GACT-PPIS outperforms the most representative models in terms of Recall, F1-measure, MCC, AUROC, and AUPRC on the benchmark test set (Test-60). Additionally, on other independent test sets (UBTest-31-6), GACT-PPIS leads in terms of Accuracy, Precision, Recall, F1-measure, MCC, AUROC, and AUPRC compared to the most representative models. It is worth noting that GACT-PPIS demonstrates excellent generalization and versatility across different test sets, showcasing good performance on multiple test sets for the same trained GACT-PPIS model.

ESM-scan-A tool to guide amino acid substitutions.

Protein structure prediction and (re)design have gone through a revolution in the last 3 years. The tremendous progress in these fields has been almost exclusively driven by readily available machine learning algorithms applied to protein folding and sequence design problems. Despite these advancements, predicting site-specific mutational effects on protein stability and function remains an unsolved problem. This is a persistent challenge, mainly because the free energy of large systems is very difficult to compute with absolute accuracy and subtle changes to protein structures are hard to capture with computational models. Here, we describe the implementation and use of ESM-Scan, which uses the ESM zero-shot predictor to scan entire protein sequences for preferential amino acid changes, thus enabling in silico deep mutational scanning experiments. We benchmark ESM-Scan on its predictive capabilities for stability and functionality of sequence changes using three publicly available datasets and proceed by experimentally testing the tool's performance on a challenging test case of a blue-light-activated diguanylate cyclase from Methylotenera species (MsLadC), where it accurately predicted the importance of a highly conserved residue in a region involved in allosteric product inhibition. Our experimental results show that the ESM-zero shot model is capable of inferring the effects of a set of amino acid substitutions in their correlation between predicted fitness and experimental results. ESM-Scan is publicly available at https://huggingface.co/spaces/thaidaev/zsp.

A novel antifungal peptide, SP1.2, from Rhodopseudomonas palustris against the rice blast pathogen.

Rice blast has a significant detrimental impact on rice yields, so developing efficient biological control technologies is an effective means for rice blast prevention and control. The GroEL protein has proven to be effective at preventing and managing the pathogenicity of rice blast. Here, we analyzed the amino acid sequence of the GroEL protein and synthesized the '60 kDa chaperonin signature' (350-373 amino acids) peptide SP1.2, which has potent antifungal activity. Notably, the SP1.2 peptide exhibited potent fungicidal activity against Magnaporthe oryzae, effectively inhibiting appressorium germination. Electron microscopy revealed that SP1.2 disrupted the fungal plasma membrane and bound to multiple bioactive phosphoinositides in vitro, triggering the production of reactive oxygen species. Furthermore, it also caused an increase in the acetylation of M. oryzae and induced autophagy in cells. The spray application of SP1.2 significantly reduced the number of disease spots caused by the fungal pathogen M. oryzae in rice, enhancing the defense response of rice plants. Field trials showed that the control effect was 64.59% after spraying SP1.2. Our study illustrates the antifungal activity of the structurally unique SP1.2 peptide against plant fungal pathogens and paves the way for the future development of this class of peptides as antifungal agents. © 2024 Society of Chemical Industry.

Bioinformatics-based Analysis of the Variability of MPOX Virus Proteins

Background: Previously restricted to remote areas of Central and Western Africa, the MPOX virus-based disease, also known as monkeypox, has now spread to more than 90 countries and has become endemic. As a consequence, the MPOX virus has become a global public health concern. Objective: The objective of this study was to conduct a computational-multiparametric study (at the genomic and proteomic levels) of the biological sequences that express the MPOX virus envelopes in order to fathom the physicochemical regularities of these proteins. Methods: Using computer programs, we determined the polarity index method (PIM) profile and protein intrinsic disorder predisposition (PIDP) for each studied protein. Results: The UniProt database was able to identify sequences similar to those of the MPOX virus expressed thanks to the computational regularities found in the virus' envelope sequences. Conclusion: The polarity index method and protein intrinsic disorder predisposition profiles could aid in elucidating the sequence-level structural regularities of the MPOX virus envelopes.

Graph-DTI: A New Model for Drug-target Interaction Prediction Based on Heterogenous Network Graph Embedding.

In this study, we aimed to develop a new end-to-end learning model called Graph-Drug-Target Interaction (DTI), which integrates various types of information in the heterogeneous network data, and to explore automatic learning of the topology-maintaining representations of drugs and targets, thereby effectively contributing to the prediction of DTI. Precise predictions of DTI can guide drug discovery and development. Most machine learning algorithms integrate multiple data sources and combine them with common embedding methods. However, the relationship between the drugs and target proteins is not well reported. Although some existing studies have used heterogeneous network graphs for DTI prediction, there are many limitations in the neighborhood information between the nodes in the heterogeneous network graphs. We studied the drug-drug interaction (DDI) and DTI from DrugBank Version 3.0, protein-protein interaction (PPI) from the human protein reference database Release 9, drug structure similarity from Morgan fingerprints of radius 2 and calculated by RDKit, and protein sequence similarity from Smith-Waterman score. Our study consists of three major components. First, various drugs and target proteins were integrated, and a heterogeneous network was established based on a series of data sets. Second, the graph neural networks-inspired graph auto-encoding method was used to extract high-order structural information from the heterogeneous networks, thereby revealing the description of nodes (drugs and proteins) and their topological neighbors. Finally, potential DTI prediction was made, and the obtained samples were sent to the classifier for secondary classification. The performance of Graph-DTI and all baseline methods was evaluated using the sums of the area under the precision-recall curve (AUPR) and the area under the receiver operating characteristic curve (AUC). The results indicated that Graph-DTI outperformed the baseline methods in both performance results. Compared with other baseline DTI prediction methods, the results showed that Graph-DTI had better prediction performance. Additionally, in this study, we effectively classified drugs corresponding to different targets and vice versa. The above findings showed that Graph-DTI provided a powerful tool for drug research, development, and repositioning. Graph- DTI can serve as a drug development and repositioning tool more effectively than previous studies that did not use heterogeneous network graph embedding.