BioKMS: A Biomedical and Space Science Knowledge Mining System Based on Knowledge Graph and Link Prediction Model

BioKMS: A Biomedical and Space Science Knowledge Mining System Based on Knowledge Graph and Link Prediction Model | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article BioKMS: A Biomedical and Space Science Knowledge Mining System Based on Knowledge Graph and Link Prediction Model Yanzhong Wen, Yuexia Han, Jianfei Sun This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5232213/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background: In recent years, Artificial Intelligence (AI) technology, represented by deep learning, has flourished, widely empowering various scientific research. At present, it is worth-noting that biological science and computer technology are developing rapidly and massive biomedical data are accumulating, accompanied by the development of advanced AI algorithms and nascent AI-enabled biomedical research opportunities. Researchers learn to combine AI with real biomedical scenarios, expecting to improve the efficacy of research. However, it is a daunting task for most biomedical researchers, who might lack the knowledge of computer science, to use AI to assist work. Therefore, an open tool integrating advanced AI methods is needed to liberate biomedical researchers from “daily chores” and assist scientific research progress. It is desirable that researchers with little knowledge of AI may still make good use of AI technology to explore scientific clues and promote scientific research. Results: Here, armed with the AI technology, we develop Biomedical Knowledge Mining System (BioKMS), a web-based biomedical knowledge mining system encompassing both biomedical and space science knowledge, which demonstrates the knowledge graph and link prediction model. More specifically, the knowledge mining refers to the use of effective methods to mine the internal relationships contained in a large number of data, providing scientific clues for biomedical studies. On the one hand, BioKMS enables researchers to explore the links between domain entity and obtain the visual presentation of the knowledge graph, intuitively capturing the domain-related information; On the other hand, researchers can carry out the link prediction with the encapsulated SOTA deep learning model, which predict the potential association, further revealing potential research objects and assisting decision-making by AI. In addition, in order to improve the timeliness and stability of the data, BioKMS provides flexible data feedback interface that researchers can submit new data, thus cementing the optimization of the subsequent algorithm model. Conclusions: In this study, we proposed a novel system to help biomedical scientists use AI to predict and discover potential biological mechanisms. In addition to serving the traditional ground biomedical research, the system integrates the knowledge graph and deep learning model of space life science field to assist the space biomedical research. The BioKMS can be accessed via this URL: http://8.130.71.195:5000/. biomedical knowledge graph space life science link prediction software system Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Background With the rapid development of information technology and the continuous breakthrough in the field of bioinformatics, the research method driven by “Big Data and AI” has become an important way to explore in the biomedical field[1]. As a frontier interdisciplinary science in the fields of life science, biology and medicine, biomedical research involves molecules, cells, tissues, organs and biological individuals, and is committed to solving related problems in life science and medicine by using engineering techniques to promote the healthy development of human life. However, biomedical research is currently facing challenges such as large amount of data, diverse data types and scattered information. For example, literature plays an important role in scientific research practice, which is an important data resource for researchers to understand the cutting-edge technology and progress. Some new ideas are often inseparable from the information mining of a large number of scientific documents. In recent years, the number of literature in the biomedical field has increased rapidly. For example, MEDLINE, an international authoritative biomedical literature database, has collected more than 36 million scientific documents, and now the literature resources are provided in electronic form and easy to access[2]. The massive knowledge it contains is crucial to promote the progress of biomedical research. At the same time, however, the biomedical research field is facing increasingly needs with regard to making good use of large-scale data, mining rich information from massive data and discovering new knowledge. The maturity of knowledge graph and deep learning technology provides a new method for efficient mining of massive data. Google first put forward the concept of knowledge map in 2012. As one of the important branches of AI technology, knowledge graph describes the concepts and relationships in the objective world in the form of triple symbols, and then constructs a network-like knowledge structure to help researchers organize data efficiently[3]. Classical knowledge graphs include YAGO, WordNet, Freebase and so on[4]. Users then can realize downstream tasks such as information mining, relational reasoning and knowledge discovery of large-scale data[4]. Deep learning aims at automatically extracting features from a large number of data and capturing potential patterns by using multi-layer neural network models. Due to different network model designs, various neural network models have been derived, such as CNN, RNN, and GNN [5]. Among them, GNN has been widely used in the analysis of graph data and achieved good results, such as node classification, link prediction, graph generation [6, 7]. Thus, it is safe to assert that the combination of knowledge graph and deep learning technology provides refreshing opportunities for biomedical researchers to “hone their scholarly disposition.” In order to accurately predict the drug target protein and reveal the drug-target interaction, Sameh K Mohamed et al. turn this prediction problem into a link prediction problem in the knowledge graph, used the biomedical knowledge base to create a knowledge graph related to drugs and their potential targets, proposing a graph embedding model to train the computer to automatically infer the target function of candidate drugs[8]; Mona Alshahrani et al. constructed a knowledge map by collecting and sorting out the data of disease and gene phenotypic relationship, and proposed a method of using feature learning to generate phenotypic representation related to entities to predict the priority order of genes related to diseases[9]; Christos Dimitrakopoulos et al. integrated different types of molecular data into a directed graph and developed a NetICS model to predict cancer factors by graph diffusion[10]. It can be seen that AI intelligent driving method has been widely used in traditional ground biomedical research. However, it is worth noting that with the continuous development and progress of space technology, outer space is gradually becoming a new frontier for scientific research. The paradigm change driven by data intelligence has gradually become the research hotspot worldwide. The United States, the European Union and other space powers and organizations have realized the importance of developing AI in space and deployed relevant strategic plans[11]. As a special branch of biomedicine, space life science is devoted to exploring life phenomena and laws in outer space environment. For example, in order to better understand the hemolysis phenomenon in microgravity environment and ensure the health status of astronauts, Ethan Waisberg developed a deep learning algorithm by using astronauts and ground simulation metadata; monitored the blood status of astronauts by non-invasive retinal imaging during space flight; and further explored the mechanism of microgravity-induced anemia[12]. Due to the particularity of the experimental environment and the lack of previous experience, many factors need to be taken into consideration when it comes to carrying out biomedical experiments in outer space. Particularly worth-noting is that AI has great potential in assisting the decision-making task of space life science research after synthesizing massive data. In this paper, we introduce an interactively-web-based knowledge mining platform that allowed biomedical researchers to use AI to assist scientific practice or scientific decision-making. Specifically, the system creates the knowledge graph in the fields of biomedicine and space life science by standardizing the data, encapsulating the SOTA deep learning model of link prediction, and designing the storage mode of the underlying data. The tasks were prioritized as follows: To demonstrate the sources, characteristics and storage methods of the data. To design a user-friendly interactive interface. To showcase the knowledge mining route for biomedicine and space life science respectively. Implementation Data acquisition and storage The knowledge graph was used as the underlying core data. Given the differences between biomedicine on the ground and life science research in space, the biomedical knowledge graph and space life science knowledge graph was sorted out and preprocessed respectively. To help fulfill the task of knowledge graph construction, SemMedDB is used. SemMedDB is a semantic predicate database extracted by SemRep, a biomedical text semantic interpreter[13, 14]. The data of SemMedDB was extracted from the scientific literature collected by MEDLINE, and the semantic triple information of the title and abstract of the literature was extracted by SemRep tool[13]. Among them, the triplet contains the concept information based on UMLS (Unified Medical Language System) and the relationship information between concepts. SemMedDB is continuously maintained by NIH(National Institutes of Health) and updated regularly. In the latest edition, the datasets covers as many as 128,903,269 pieces of semantic predicate information, involving 68 association types and 127 biomedical entity types. The SemMedDB triplet data was downloaded from NIH official website. Through the entity type designation and entity unique identification with experts' participation, we cleaned and redundantly preprocessed the original data, and obtained standardized triplet information and entity information. Moreover, we use preprocessed triples to construct a knowledge graph, with entities as nodes and relationships as edges, which contains about 269,951 biomedical entities and 16,138,471 relationships. In order to ensure the query efficiency, we only keep isomorphic edges in the construction of knowledge graph. This indicates that, for two entities with multiple edges (relationships), one edge was kept and the detailed relation information was stored as relational data. As the interdisciplinary study that encompasses space science and life science, space life science is committed to exploring the law of life from the influence of space factors and feeding back the biomedical research on the ground. Due to the complexity and particularity of the field, the relevant knowledge sources in the field of space life science are extensive, further exacerbated by the varying storage forms. This fragmentation of knowledge helps clarify why the knowledge graph in this field is relatively lacking. To help researchers quickly catch a snapshot of the knowledge and the prospect in the field, the construction of the knowledge graph seems timely and necessary. Therefore, the scientific literature related to space life science in WOS(Web Of Science) and the experimental information released by NASA was collected respectively, and the software entitled SemRep was applied to extract entities and relationships, and finally, the independent construction of Space-Life-Science Knowledge Graph (SLS-KG) was constructed, involving 6,171 entities and 23,885 relationships. With respect to the graph data, we use the graph database Neo4j Desktop 1.5.8 to store and manage, as shown in Fig. 1, and use the Neo4j API to query. For normalized entity and relationship information, we use relational database MySQL to store, so as to achieve higher query efficiency. In addition, in order to realize the function of entity query based on semantic similarity, BioBERT[15], a biomedical pre-training language model, was used to transform entity names into vector representations and the vector database titled Faiss[16], was used to construct indexes, further realizing fast retrieval based on vector similarity. System design and implementation The BioKMS web server, deployed in the windows 10 environment, uses the Python Web framework flask 2.3.2 and the front-end framework Vue3. With regard to the front-end interface, HTML, CSS, Javascript was used to design the page layout and response. The Element-plus front-end component library was used to render page buttons, tables and other elements. D3 library was used to complete the rendering of the graph. The Axios library undertook the task of data transmission between the front and back end. The back-end link prediction model was developed by the team members in charge of algorithm research based on Graph Neural Networks[17] and pre-training model. In this work, the trained model is encapsulated by deep learning frameworks such as pytorch, DGL, network, further ensuring that the system can be used for interactive use in a user-friendly fashion. The BioKMS sysem is freely available at http://8.130.71.195:5000/.  Results A summary of BioKMS BioKMS is a biomedical knowledge mining system with knowledge graph and link prediction model as its core, aiming at providing convenience for biomedical researchers as well as space science scientists by integrating the AI technology. The massive accumulation of scientific literature and the rapid growth rate of literature have made the case particularly daunting for researchers to catch the panorama of the current research progress. The system equips researchers with sufficient tools to navigate the vast amount of information embedded in the literature. Specifically, the operational flow of the construction of the system lies in threefold: first, the knowledge graph from the literature text was selected as the underlying data, second, a model based on the domain knowledge graph was created, and finally, the deep learning model was trained to realize the link prediction of domain entities, which helps researchers explore the potential research direction. With regard to the specific characteristics of the system, knowledge graph provides opportunities for text mining and its natural network structure has advantages in mining the relationship between concepts. Particularly noteworthy is that the system provides a simplistic and user-friendly platform for researchers to conveniently access knowledge graph and link prediction models, ultimately boosting their scientific research endeavors. In addition, BioKMS also achieves the functionality of constant update of its database by requesting users to provide feedback such as new relationships and findings found in the recently published research, thus improving the richness and up-to-date nature of the database and undergirding the performance of the model. User interface On the home page, BioKMS consists of two parts: the sub-graph rendering of biomedical knowledge graph and the interface of functional modules. BioKMS includes five functional modules: biomedical entity query, biomedical pathway retrieval, biomedical link prediction, space life science knowledge mining and data feedback. Among them, space life science knowledge mining involves two sections which are knowledge graph retrieval and link prediction. The biomedical entity query page (Fig. 2C) queries the entity information based on semantic similarity by obtaining the entity name to be retrieved by the user and the expected number of returned results, rendering it in the form of a table. In the initial state, the biomedical pathway (Fig. 2D) acquires the name of the head and tail entity; process the type of the path entity and the expected path length by the user; queries the path between the head and tail entities based on the knowledge graph; and renders the results accordingly. Biomedical link prediction (Fig. 2E) demonstrates the entity name to be predicted by the user and the expected topk parameters for prediction calculation. With regard to the prediction result, the user can track the interested research direction in accordance with the entity type. It is worth mentioning that the space life science knowledge mining module takes a sub-homepage (Fig. 3A) as the entrance, presenting the sub-graph rendering and functional units of the space life science knowledge graph by following a three-step pathway. First, the knowledge graph query page (Fig. 3B) captures the name of the space life science entity that the user intends to retrieve, and returns to the visual presentation of the network; second, the link prediction page (Fig. 3C) obtains the name of the space life science entity to be predicted by the user for prediction calculation; third, the user can process the result according to the entity type for the prediction result accordingly. One hallmark of the system is the exhibition of the data feedback page. The data feedback page consists of three parts, namely, new entity submission, new relationship submission and system feedback. The first two parts request users to add new data, and set up feedback interfaces in biomedical and space life sciences respectively (Fig. 4A&B); System feedback (Fig. 4C) allows users to give real time science data to help consolidate the existing database and may play a pivotal role in the optimization of the system performance. Case study In order to shed illuminating light on the distinctive nature of our system in empowering both biomedicine (on the ground) and space life science (in space), this section presents two case studies that demonstrate the potential of our system to accomplish the knowledge mining in biomedical and space life science fields. The first case study has carried out entity query, graph retrieval, link prediction calculation and pathway mining between osteoporosis and nerves, which are of great concern to researchers in the field of biomedicine. The second case study presents the retrieval of knowledge graph and the calculation of link prediction under the background of space life science. Case study 1 — Osteoporosis In order to better showcase the effect of this system in practical use, this case is aimed at osteoporosis, which is currently studied extensively by researchers from around the world[18]. Firstly, the entity information of osteoporosis is queried on the biomedical entity query page (Fig. 5A), and its associated visual atlas is obtained (Fig. 5B). Secondly, osteoporosis is regarded as a quasi-prediction entity on the biomedical link prediction page, and the topk parameter is set to 20. After the preliminary prediction result is obtained, the entity type is defined to further process the result, and finally the prediction result is visually rendered (Fig. 5C). Given the voluminous research on the mechanism between osteoporosis and nerve, we set the head and tail entities as nerve and osteoporosis in the biomedical pathway retrieval page, and set the pathway length to 3 to search. Different colors in the rendering results represent varying types of entities, which helps users to intuitively understand the enlightenment contained in the graph (Fig. 5D). Case study 2 — Gravity In the scenario of space life science, we chose ‘microgravity’ with scene characteristics as a demonstration case. First, graph visualization information of microgravity on the knowledge graph query page is shown (Fig. 6A); Secondly, the link prediction of ‘microgravity’ is calculated on the link prediction page and the entity type is further selected as ‘Disease’ after the preliminary results. Later on, the diseases that are potentially related to microgravity are observed. Finally, the prediction results are visually rendered (Fig. 6B). Discussion Requirement analysis In recent years, the rapid development and extensive empowerment of AI technology have attracted researchers worldwide to carry out research on the migration and application of cutting-edge computer technology, followed by advanced knowledge maps and efficient knowledge organization forms. Given the rising of the space science research, there seems to be an urgent call for the integration of the research on the ground and in space. This gap seemingly could be addressed with the skyrocketing development of Artificial Intelligence technology by mining the domain knowledge. On the one hand, AI helps resolve the prohibitively technical cost of learning to learn new fields; On the other hand, the mastery of AI technology for experts in another field is a daunting challenge. Therefore, integrating diversified knowledge mining functions based on AI into an open software system can well alleviate the research problems of researchers and help solve the knowledge mining needs of researchers. The system, titled BioKMS, exhibits good characteristics such as data query, model calculation and result visualization. The system is designed with a particular focus on biomedical and space life science. Potential users may use domain knowledge map to query domain entities and related networks between entities. Moreover, the association prediction model based on deep learning is used to calculate the association prediction of entities and predict the potential research targets. As shown in Fig. 7, first, the user searches the entity to view the basic information and relationship network of the target entity; Secondly, the target entity is calculated by association prediction, and the preliminary prediction results are analyzed twice (filtered by type or keyword) in regard to interested entities. The user may visually observe the association prediction results and export and save them in the personal computer. Third, on the basis of the valuable results obtained, users can further carry out biological experiments for verification, further corroborating the authenticity of the data and submitting the feedback accordingly. Benefits and potential BioKMS galvanizes AI-assisted biomedical and space life science research in an easy-to-access form. It is important to note that, intriguingly and surprisingly, researchers with little knowledge of AI may still make good use of AI technology to explore scientific clues and promote scientific research. Following the paradigm shown in Fig. 7, BioKMS can provide direction of future research through knowledge graph query and deep learning model prediction; inspire scientists to delve into exciting research both on the ground and in space; and the newly fed knowledge into the database by the scientists is an equally impressive hallmark of the system. Conclusions We have developed a Biomedical Knowledge Mining System, BioKMS, which integrates AI method in order to solve the pressing needs of biomedical researchers and space science scientists, ultimately expecting to use AI to mine domain knowledge, inspire scientific research ideas, and thus drive scientific research. With user-friendly interactive interface and diversified functions, BioKMS fully takes into account the needs and expectations of researchers in real scientific research scenes. In the traditional biomedical field, BioKMS is a novel knowledge mining service platform that combines knowledge graph and deep learning model. To add to the novelty of the system, BioKMS innovatively couples knowledge graph in the space life science to realize deep mining of knowledge. With the continuous enrichment of data and the continuous optimization of AI model, it is our conviction that BioKMS can shed substantial light on the future scientific research both on the ground and in space. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Availability of data and materials Data sharing is not applicable to this article as no datasets were generated or analysed during the current study. Funding Not applicable. Competing interests The authors declare no competing interests. Authors' contributions All Authors have made significant contributions to the work. YZ. Wen: conceptualization, investigation, data curation, methodology, supervision, writing-original draft, review and editing. YX. Han: conceptualization, investigation, data curation, methodology, supervision, writing-original draft, review and editing. JF. Sun: conceptualization, investigation, data curation, methodology, supervision, writing-original draft, review and editing. All authors reviewed the manuscript and approved the final version. Acknowledgements Not applicable. Authors' information YZ. Wen (Email: [email protected] ) is a graduate student in the School of Biological Sciences and Medical Engineering of Southeast University and the Jiangsu Key Laboratory for Biomaterials and Devices. Dr. YX. Han (Email: [email protected] ) is a faculty advisor to graduate student in the School of Biological Sciences and Medical Engineering of Southeast University and the Jiangsu Key Laboratory for Biomaterials and Devices. Prof. Dr. JF. Sun (Email: [email protected] ) is a Director in the School of Biological Sciences and Medical Engineering of Southeast University and the Jiangsu Key Laboratory for Biomaterials and Devices. Availability and requirements Project name: BioKMS Project home page: http://8.130.71.195:5000/ Operating system(s): Windows 10/11 Programming language: Python Other requirements: Python 3.10.9 or higher, Java 17.0.4 or higher, Neo4j Desktop 1.5.8 etc. License: GNU GPL Any restrictions to use by non-academics: licence needed References Islam MT, Xing L. A data-driven dimensionality-reduction algorithm for the exploration of patterns in biomedical data. NAT BIOMED ENG. 2021;5(6):624-35. Zhao S, Su C, Lu Z, Wang F. Recent advances in biomedical literature mining. BRIEF BIOINFORM. 2021;22(3):a57. Fensel D, Şimşek U, Angele K, Huaman E, Kärle E, Panasiuk O, et al. Introduction: what is a knowledge graph? Knowledge graphs: Methodology, tools and selected use cases. 2020:1-10. Chen X, Jia S, Xiang Y. A review: Knowledge reasoning over knowledge graph. EXPERT SYST APPL. 2020;141:112948. Dong S, Wang P, Abbas K. A survey on deep learning and its applications. COMPUT SCI REV. 2021;40:100379. Zhang X, Liang L, Liu L, Tang M. Graph neural networks and their current applications in bioinformatics. FRONT GENET. 2021;12:690049. Zhang M. Graph neural networks: link prediction. Graph Neural Networks: Foundations, Frontiers, and Applications. 2022:195-223. Mohamed SK, Nováček V, Nounu A. Discovering protein drug targets using knowledge graph embeddings. BIOINFORMATICS. 2020;36(2):603-10. Alshahrani M, Hoehndorf R. Semantic disease gene embeddings (smudge): phenotype-based disease gene prioritization without phenotypes. BIOINFORMATICS. 2018;34(17):i901-7. Dimitrakopoulos C, Hindupur SK, Häfliger L, Behr J, Montazeri H, Hall MN, et al. Network-based integration of multi-omics data for prioritizing cancer genes. BIOINFORMATICS. 2018;34(14):2441-8. Nita G, Georgoulis M, Kitiashvili I, Sadykov V, Camporeale E, Kosovichev A, et al. Machine learning in heliophysics and space weather forecasting: a white paper of findings and recommendations. arXiv preprint arXiv:2006.12224. 2020. Waisberg E, Ong J, Zaman N, Kamran SA, Lee AG, Tavakkoli A. A non-invasive approach to monitor anemia during long-duration spaceflight with retinal fundus images and deep learning. LIFE SCI SPACE RES. 2022;33:69-71. Kilicoglu H, Shin D, Fiszman M, Rosemblat G, Rindflesch TC. SemMedDB: a PubMed-scale repository of biomedical semantic predications. BIOINFORMATICS. 2012;28(23):3158-60. Arnold P, Rahm E. SemRep: A repository for semantic mapping. 2015. Lee J, Yoon W, Kim S, Kim D, Kim S, So CH, et al. BioBERT: a pre-trained biomedical language representation model for biomedical text mining. BIOINFORMATICS. 2020;36(4):1234-40. Douze M, Guzhva A, Deng C, Johnson J, Szilvasy G, Mazaré P, et al. The faiss library. arXiv preprint arXiv:2401.08281. 2024. Hamilton W, Ying Z, Leskovec J. Inductive representation learning on large graphs. Advances in neural information processing systems. 2017;30. Rachner TD, Khosla S, Hofbauer LC. Osteoporosis: now and the future. The Lancet. 2011;377(9773):1276-87. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5232213","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":379592474,"identity":"506d50a9-9397-4279-8037-46ba6543750f","order_by":0,"name":"Yanzhong Wen","email":"","orcid":"","institution":"Jiangsu Key Laboratory for Biomaterials and Devices, School of Biological Sciences and Medical Engineering, Southeast University, 211189 Nanjing, China.","correspondingAuthor":false,"prefix":"","firstName":"Yanzhong","middleName":"","lastName":"Wen","suffix":""},{"id":379592475,"identity":"de116770-baa7-43df-9307-1f460a8f2856","order_by":1,"name":"Yuexia Han","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAs0lEQVRIiWNgGAWjYLCCBAYb0rWkkW7PYRLUGtxIPibxoOa8vPyMBMYPPxjs8ojQkpZskHDstuGGGwnMkj0MycVEaMkxfJDAdjvBQCKBQZqB4UBiA2Et+R8OJPw7lwB0GPNvIrXkMD5IbDuQwHAjgY04WyTPPDM2SOxLNtxw5mGbZY9BMmEtfMeTn0n++GYnL9+efPjGjwo7wloUDsCZjEDFBoTUA4E8QUNHwSgYBaNgFAAAGu09+dZN6XEAAAAASUVORK5CYII=","orcid":"","institution":"Center for Space Application Engineering and Technology, Chinese Academy of Sciences, 100094 Beijing, China.","correspondingAuthor":true,"prefix":"","firstName":"Yuexia","middleName":"","lastName":"Han","suffix":""},{"id":379592476,"identity":"a83f0659-99ec-4dc0-8f0b-8bbbc12f5dc8","order_by":2,"name":"Jianfei Sun","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA1UlEQVRIiWNgGAWjYPCCAzz8EAYzCVokG0jVwmBwgFgt8jNyD3/48euOjPH5w8ckGCqsExvYzx7Aq8XgRl6CYW/fMx6zG2lpEgxn0hMbePIS8GuRyDFI4O05DNTCYybB2HY4sUGCx4CAw3IMDv4FajHuP/9NgvEfEVoYbuQYNvP8OAxUlsMmwdhAhBaDM2+MmWUbDvNI3Egztkg4lm7cxpNDwGHtOcYf3/w5bM/ff/jhjQ811rL97GcIOAwEGNugjAQgZiOsHgT+EKdsFIyCUTAKRigAAO7SRb/2IU8oAAAAAElFTkSuQmCC","orcid":"","institution":"Jiangsu Key Laboratory for Biomaterials and Devices, School of Biological Sciences and Medical Engineering, Southeast University, 211189 Nanjing, China.","correspondingAuthor":true,"prefix":"","firstName":"Jianfei","middleName":"","lastName":"Sun","suffix":""}],"badges":[],"createdAt":"2024-10-09 11:38:19","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5232213/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5232213/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":69361983,"identity":"2ac2d404-faea-4bea-89e2-202b402b9054","added_by":"auto","created_at":"2024-11-19 14:26:49","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":383854,"visible":true,"origin":"","legend":"\u003cp\u003eOverview of the Neo4j Database Management System. A Neo4j Desktop instance management interface. B Partial visualization of biomedical knowledge graph in Neo4j. C Partial visualization of SLS-KG in Neo4j.\u003c/p\u003e","description":"","filename":"Fig1.png","url":"https://assets-eu.researchsquare.com/files/rs-5232213/v1/a7ec561d3183396ccfde260b.png"},{"id":69361982,"identity":"ef081f5a-0738-4576-bea9-969068bb2fde","added_by":"auto","created_at":"2024-11-19 14:26:49","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":207114,"visible":true,"origin":"","legend":"\u003cp\u003eBioKMS content and web interface. A Home page, including a sub-graph rendering of biomedical knowledge graph. B Functional module overview page, including five core functional module jump links. C Biomedical entity query page, consisted of a user input area and a result presentation area. D Biomedical pathway retrieval page, returning the pathway visualization. E Biomedical link prediction page, executing link compute while obtaining the entity name to be predicted.\u003c/p\u003e","description":"","filename":"Fig2.png","url":"https://assets-eu.researchsquare.com/files/rs-5232213/v1/9cabe7ed7e455bddd1b4d061.png"},{"id":69361986,"identity":"4193c885-6324-4cc7-8fcd-ae9440762deb","added_by":"auto","created_at":"2024-11-19 14:26:49","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":643641,"visible":true,"origin":"","legend":"\u003cp\u003eSpace life science module content. A Sub-home page, including a sub-graph rendering of space-life-science knowledge graph. B Knowledge graph query page, obtaining the interested entity name and returning the results by graph. C link prediction page, executing link compute and returning the prediction results.\u003c/p\u003e","description":"","filename":"Fig3.png","url":"https://assets-eu.researchsquare.com/files/rs-5232213/v1/a75a357b319ebb4746ef4be9.png"},{"id":69362497,"identity":"1cfbcfc9-0cd4-44ab-a3db-22b949346f3c","added_by":"auto","created_at":"2024-11-19 14:34:49","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":36471,"visible":true,"origin":"","legend":"\u003cp\u003eData Feedback module content. A New entity submission page. B New relationship submission. C System feedback page.\u003c/p\u003e","description":"","filename":"Fig4.png","url":"https://assets-eu.researchsquare.com/files/rs-5232213/v1/d9e97c627dc6e6c68bbcd278.png"},{"id":69361988,"identity":"ddee1b3c-17ea-4b7c-bc0d-36c23fc0a58f","added_by":"auto","created_at":"2024-11-19 14:26:49","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":299395,"visible":true,"origin":"","legend":"\u003cp\u003eDisplay of case study 1. An Entity query results, using ‘Osteoporosis’ as the target entity. B The graph of the ‘Osteoporosis’ based on knowledge graph retrieval. C Link prediction results, setting topk to 20 and entity type to ‘Anatomy’\u0026amp;’Genes \u0026amp; Molecular Sequences’, then the prediction results are visualized. D Biomedical pathway retrieval results, setting head entity to ‘Nerve’, tail entity to ‘Osteoporosis’ and pathway length to ‘3’.\u003c/p\u003e","description":"","filename":"Fig5.png","url":"https://assets-eu.researchsquare.com/files/rs-5232213/v1/83eaef544288b9b3811b20b3.png"},{"id":69361985,"identity":"9d96f6c7-706b-4e32-a0df-6ae8d264c338","added_by":"auto","created_at":"2024-11-19 14:26:49","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":326017,"visible":true,"origin":"","legend":"\u003cp\u003eDisplay of case study 2. A Knowledge graph query results, using ‘Microgravity’ as the target entity. B Link prediction results, setting entity type to ‘Disease_Syndrome’, then the prediction results are visualized.\u003c/p\u003e","description":"","filename":"Fig6.png","url":"https://assets-eu.researchsquare.com/files/rs-5232213/v1/32153357356f0e30129f47dc.png"},{"id":69361987,"identity":"3513e84e-18bd-43e9-bb9b-3889342f6c4a","added_by":"auto","created_at":"2024-11-19 14:26:49","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":187580,"visible":true,"origin":"","legend":"\u003cp\u003eThe overall process of users using the system.\u003c/p\u003e","description":"","filename":"Fig7.png","url":"https://assets-eu.researchsquare.com/files/rs-5232213/v1/385c00ad1b39274649648b38.png"},{"id":69362499,"identity":"a432376a-1929-4c78-b03e-910fff656b60","added_by":"auto","created_at":"2024-11-19 14:34:54","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2242185,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5232213/v1/8e9a86d5-4494-489f-9e6c-4f7cb84223ee.pdf"}],"financialInterests":"","formattedTitle":"BioKMS: A Biomedical and Space Science Knowledge Mining System Based on Knowledge Graph and Link Prediction Model","fulltext":[{"header":"Background","content":"\u003cp\u003eWith the rapid development of information technology and the continuous breakthrough in the field of bioinformatics, the research method driven by \u0026ldquo;Big Data and AI\u0026rdquo; has become an important way to explore in the biomedical field[1]. As a frontier interdisciplinary science in the fields of life science, biology and medicine, biomedical research involves molecules, cells, tissues, organs and biological individuals, and is committed to solving related problems in life science and medicine by using engineering techniques to promote the healthy development of human life. However, biomedical research is currently facing challenges such as large amount of data, diverse data types and scattered information. For example, literature plays an important role in scientific research practice, which is an important data resource for researchers to understand the cutting-edge technology and progress. Some new ideas are often inseparable\u0026nbsp;\u003c/p\u003e\n\u003cp\u003efrom the information mining of a large number of scientific documents. In recent years, the number of literature in the biomedical field has increased rapidly. For example, MEDLINE, an international authoritative biomedical literature database, has collected more than 36 million scientific documents, and now the literature resources are provided in electronic form and easy to access[2]. The massive knowledge it contains is crucial to promote the progress of biomedical research. At the same time, however, the biomedical research field is facing increasingly needs with regard to making good use of large-scale data, mining rich information from massive data and discovering new knowledge.\u003c/p\u003e\n\u003cp\u003eThe maturity of knowledge graph and deep learning technology provides a new method for efficient mining of massive data. Google first put forward the concept of knowledge map in 2012. As one of the important branches of AI technology, knowledge graph describes the concepts and relationships in the objective world in the form of triple symbols, and then constructs a network-like knowledge structure to help researchers organize data efficiently[3]. Classical knowledge graphs include YAGO, WordNet, Freebase and so on[4]. \u0026nbsp; Users then can realize downstream tasks such as information mining, relational reasoning and knowledge discovery of large-scale data[4]. Deep learning aims at automatically extracting features from a large number of data and capturing potential patterns by using multi-layer neural network models. Due to different network model designs, various neural network models have been derived, such as CNN, RNN, and GNN\u0026nbsp;[5]. Among them, GNN has been widely used in the analysis of graph data and achieved good results, such as node classification, link prediction, graph generation\u0026nbsp;[6, 7]. Thus, it is safe to assert that the combination of knowledge graph and deep learning technology provides refreshing opportunities for biomedical researchers to \u0026ldquo;hone their scholarly disposition.\u0026rdquo; In order to accurately predict the drug target protein and reveal the drug-target interaction, Sameh K Mohamed et al. turn this prediction problem into a link prediction problem in the knowledge graph, used the biomedical knowledge base to create a knowledge graph related to drugs and their potential targets, proposing a graph embedding model to train the computer to automatically infer the target function of candidate drugs[8]; Mona Alshahrani et al. constructed a knowledge map by collecting and sorting out the data of disease and gene phenotypic relationship, and proposed a method of using feature learning to generate phenotypic representation related to entities to predict the priority order of genes related to diseases[9]; Christos Dimitrakopoulos et al. integrated different types of molecular data into a directed graph and developed a NetICS model to predict cancer factors by graph diffusion[10].\u003c/p\u003e\n\u003cp\u003eIt can be seen that AI intelligent driving method has been widely used in traditional ground biomedical research. However, it is worth noting that with the continuous development and progress of space technology, outer space is gradually becoming a new frontier for scientific research. The paradigm change driven by data intelligence has gradually become the research hotspot worldwide. The United States, the European Union and other space powers and organizations have realized the importance of developing AI in space and deployed relevant strategic plans[11]. As a special branch of biomedicine, space life science is devoted to exploring life phenomena and laws in outer space environment. For example, in order to better understand the hemolysis phenomenon in microgravity environment and ensure the health status of astronauts, Ethan Waisberg developed a deep learning algorithm by using astronauts and ground simulation metadata; monitored the blood status of astronauts by non-invasive retinal imaging during space flight; and further explored the mechanism of microgravity-induced anemia[12]. Due to the particularity of the experimental environment and the lack of previous experience, many factors need to be taken into consideration when it comes to carrying out biomedical experiments in outer space. Particularly worth-noting is that AI has great potential in assisting the decision-making task of space life science research after synthesizing massive data.\u003c/p\u003e\n\u003cp\u003eIn this paper, we introduce an interactively-web-based knowledge mining platform that allowed biomedical researchers to use AI to assist scientific practice or scientific decision-making. Specifically, the system creates the knowledge graph in the fields of biomedicine and space life science by standardizing the data, encapsulating the SOTA deep learning model of link prediction, and designing the storage mode of the underlying data.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe tasks were prioritized as follows:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eTo demonstrate the sources, characteristics and storage methods of the data.\u003c/li\u003e\n \u003cli\u003eTo design a user-friendly interactive interface.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eTo showcase the knowledge mining route for biomedicine and space life science respectively.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3\u003eImplementation\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003eData acquisition and storage\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe knowledge graph was used as the underlying core data. Given the differences between \u0026nbsp; biomedicine on the ground and life science research in space, the biomedical knowledge graph and space life science knowledge graph was sorted out and preprocessed respectively. To help fulfill the task of knowledge graph construction, SemMedDB is used. SemMedDB is a semantic predicate database extracted by SemRep, a biomedical text semantic interpreter[13, 14]. The data of SemMedDB was \u0026nbsp;extracted from the scientific literature collected by MEDLINE, and the semantic triple information of the title and abstract of the literature was extracted by SemRep tool[13]. Among them, the triplet contains the concept information based on UMLS (Unified Medical Language System) and the relationship information between concepts. SemMedDB is continuously maintained by NIH(National Institutes of Health) and updated regularly. In the latest edition, the datasets covers as many as 128,903,269 pieces of semantic predicate information, involving 68 association types and 127 biomedical entity types. The \u0026nbsp; SemMedDB triplet data was downloaded from NIH official website. Through the entity type designation and entity unique identification with experts\u0026apos; participation, we cleaned and redundantly preprocessed the original data, and obtained standardized triplet information and entity information. Moreover, we use preprocessed triples to construct a knowledge graph, with entities as nodes and relationships as edges, which contains about 269,951 biomedical entities and 16,138,471 relationships. In order to ensure the query efficiency, we only keep isomorphic edges in the construction of knowledge graph. This indicates that, for two entities with multiple edges (relationships), one edge was kept and the detailed relation information was stored as relational data. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAs the interdisciplinary study that encompasses space science and life science, space life science is committed to exploring the law of life from the influence of space factors and feeding back the biomedical research on the ground. Due to the complexity and particularity of the field, the relevant knowledge sources in the field of space life science are extensive, further exacerbated by the varying storage forms. This fragmentation of knowledge helps clarify why the knowledge graph in this field is relatively lacking. To help researchers quickly catch a snapshot of the knowledge and the prospect in the field, the construction of the knowledge graph seems timely and necessary.\u003c/p\u003e\n\u003cp\u003eTherefore, the scientific literature related to space life science in WOS(Web Of Science) and the experimental information released by NASA was collected respectively, and the software entitled SemRep was applied to extract entities and relationships, and finally, the independent construction of Space-Life-Science Knowledge Graph (SLS-KG) was constructed, involving 6,171 entities and 23,885 relationships.\u003c/p\u003e\n\u003cp\u003eWith respect to the graph data, we use the graph database Neo4j Desktop 1.5.8 to store and manage, as shown in Fig. 1,\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eand use the Neo4j API to query. For normalized entity and relationship information, we use relational database MySQL to store, so as to achieve higher query efficiency. In addition, in order to realize the function of entity query based on semantic similarity, BioBERT[15], a biomedical pre-training language model, was used to transform entity names into vector representations and the vector database titled Faiss[16], was used to construct indexes, further realizing fast retrieval based on vector similarity.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSystem design and implementation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe BioKMS web server, deployed in the windows 10 environment, uses the Python Web framework flask 2.3.2 and the front-end framework Vue3. With regard to the front-end interface, HTML, CSS, Javascript was used to design the page layout and response. The Element-plus front-end component library was used to render page buttons, tables and other elements. D3 library was used to complete the rendering of the graph. The Axios library undertook the task of data transmission between the front and back end.\u003c/p\u003e\n\u003cp\u003eThe back-end link prediction model was developed by the team members in charge of algorithm research based on Graph Neural Networks[17] and pre-training model. In this work, the trained model is encapsulated by deep learning frameworks such as pytorch, DGL, network, further ensuring that the system can be used for interactive use in a user-friendly fashion. The BioKMS sysem is freely available at \u0026nbsp;http://8.130.71.195:5000/.\u0026amp;nbsp;\u0026nbsp;\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eA summary of BioKMS\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBioKMS is a biomedical knowledge mining system with knowledge graph and link prediction model as its core, aiming at providing convenience for biomedical researchers as well as space science scientists by integrating the AI technology. The massive accumulation of scientific literature and the rapid growth rate of literature have made the case particularly daunting for researchers to catch the panorama of the current research progress. The system equips researchers with sufficient tools to navigate the vast amount of information embedded in the literature. Specifically, the operational flow of the construction of the system lies in threefold: first, the knowledge graph from the literature text was selected as the underlying data, second, a model based on the domain knowledge graph was created, and finally, the deep learning model was trained to realize the link prediction of domain entities, which helps researchers explore the potential research direction. With regard to the specific characteristics of the system, knowledge graph provides opportunities for text mining and its natural network structure has advantages in mining the relationship between concepts. Particularly noteworthy is that the system provides a simplistic and user-friendly platform for researchers to conveniently access knowledge graph and link prediction models, ultimately boosting their scientific research endeavors. In addition, BioKMS also achieves the functionality of constant update of its database by requesting users to provide feedback such as new relationships and findings found in the recently published research, thus improving the richness and up-to-date nature of the database and undergirding the performance of the model.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eUser interface\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOn the home page, BioKMS consists of two parts: the sub-graph rendering of biomedical knowledge graph and the interface of functional modules. BioKMS includes five functional modules: biomedical entity query, biomedical pathway retrieval, biomedical link prediction, space life science knowledge mining and data feedback. Among them, space life science knowledge mining involves two sections which are knowledge graph retrieval and link prediction. \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe biomedical entity query page (Fig. 2C) queries the entity information based on semantic similarity by obtaining the entity name to be retrieved by the user and the expected number of returned results, rendering it in the form of a table. In the initial state, the biomedical pathway (Fig. 2D) acquires the name of the head and tail entity; process the type of the path entity and the expected path length by the user; queries the path between the head and tail entities based on the knowledge graph; and renders the results accordingly. Biomedical link prediction (Fig. 2E) demonstrates the entity name to be predicted by the user and the expected topk parameters for prediction calculation. With regard to the prediction result, the user can track the interested research direction in accordance with the entity type.\u003c/p\u003e\n\u003cp\u003eIt is worth mentioning that the space life science knowledge mining module takes a sub-homepage (Fig. 3A) as the entrance, presenting the sub-graph rendering and functional units of the space life science knowledge graph by following a three-step pathway. First, the knowledge graph query page (Fig. 3B) captures the name of the space life science entity that the user intends to retrieve, and returns to the visual presentation of the network; second, the link prediction page (Fig. 3C) obtains the name of the space life science entity to be predicted by the user for prediction calculation; third, the user can process the result according to the entity type for the prediction result accordingly.\u003c/p\u003e\n\u003cp\u003eOne hallmark of the system is the exhibition of the data feedback page. The data feedback page consists of three parts, namely, new entity submission, new relationship submission and system feedback. The first two parts request users to add new data, and set up feedback interfaces in biomedical and space life sciences respectively (Fig. 4A\u0026amp;B); System feedback (Fig. 4C) allows users to give real time science data to help consolidate the existing database and may play a pivotal role in the optimization of the system performance. \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCase study\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn order to shed illuminating light on the distinctive nature of our system in empowering both biomedicine (on the ground) and space life science (in space), this section presents two case studies that demonstrate the potential of our system to accomplish the knowledge mining in biomedical and space life science fields. The first case study has carried out entity query, graph retrieval, link prediction calculation and pathway mining between osteoporosis and nerves, which are of great concern to researchers in the field of biomedicine. The second case study presents the retrieval of knowledge graph and the calculation of link prediction under the background of space life science.\u003c/p\u003e\n\u003cp\u003eCase study 1 \u0026mdash; Osteoporosis\u003c/p\u003e\n\u003cp\u003eIn order to better showcase the effect of this system in practical use, this case is aimed at osteoporosis, which is currently studied extensively by researchers from around the world[18]. Firstly, the entity information of osteoporosis is queried on the biomedical entity query page (Fig. 5A), and its associated visual atlas is obtained (Fig. 5B). Secondly, osteoporosis is regarded as a quasi-prediction entity on the biomedical link prediction page, and the topk parameter is set to 20. After the preliminary prediction result is obtained, the entity type is defined to further process the result, and finally the prediction result is visually rendered (Fig. 5C). Given the voluminous research on the mechanism between osteoporosis and nerve, we set the head and tail entities as nerve and osteoporosis in the biomedical pathway retrieval page, and set the pathway length to 3 to search. Different colors in the rendering results represent varying types of entities, which helps users to intuitively understand the enlightenment contained in the graph (Fig. 5D).\u003c/p\u003e\n\u003cp\u003eCase study 2 \u0026mdash; Gravity\u003c/p\u003e\n\u003cp\u003eIn the scenario of space life science, we chose \u0026lsquo;microgravity\u0026rsquo; with scene characteristics as a demonstration case. First, graph visualization information of microgravity on the knowledge graph query page is shown (Fig. 6A); Secondly, the link prediction of \u0026lsquo;microgravity\u0026rsquo; is calculated on the link prediction page and the entity type is further selected as \u0026lsquo;Disease\u0026rsquo; after the preliminary results. Later on, the diseases that are potentially related to microgravity are observed. \u0026nbsp;Finally, the prediction results are visually rendered (Fig. 6B).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003e\u003cstrong\u003eRequirement analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn recent years, the rapid development and extensive empowerment of AI technology have attracted researchers worldwide to carry out research on the migration and application of cutting-edge computer technology, followed by advanced knowledge maps and efficient knowledge organization forms. Given the rising of the space science research, there seems to be an urgent call for the integration of the research on the ground and in space. This gap seemingly could be addressed with the skyrocketing development of Artificial Intelligence technology by mining the domain knowledge.\u003c/p\u003e\n\u003cp\u003eOn the one hand, AI helps resolve the prohibitively technical cost of learning to learn new fields; On the other hand, the mastery of AI technology for experts in another field is a daunting challenge. Therefore, integrating diversified knowledge mining functions based on AI into an open software system can well alleviate the research problems of researchers and help solve the knowledge mining needs of researchers.\u003c/p\u003e\n\u003cp\u003eThe system, titled BioKMS, exhibits good characteristics such as data query, model calculation and result visualization. The system is designed with a particular focus on biomedical and space life science. Potential users may use domain knowledge map to query domain entities and related networks between entities. Moreover, the association prediction model based on deep learning is used to calculate the association prediction of entities and predict the potential research targets. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAs shown in Fig. 7, first, the user searches the entity to view the basic information and relationship network of the target entity; Secondly, the target entity is calculated by association prediction, and the preliminary prediction results are analyzed twice (filtered by type or keyword) in regard to interested entities. The user may visually observe the association prediction results and export and save them in the personal computer. Third, on the basis of the valuable results obtained, users can further carry out biological experiments for verification, further corroborating the authenticity of the data and submitting the feedback accordingly.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBenefits and potential\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBioKMS galvanizes AI-assisted biomedical and space life science research in an easy-to-access form. It is important to note that, intriguingly and surprisingly, researchers with little knowledge of AI may still make good use of AI technology to explore scientific clues and promote scientific research. Following the paradigm shown in Fig. 7, BioKMS can provide direction of future research through knowledge graph query and deep learning model prediction; inspire scientists to delve into exciting research both on the ground and in space; and the newly fed knowledge into the database by the scientists is an equally impressive hallmark of the system. \u0026nbsp;\u0026nbsp;\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eWe have developed a Biomedical Knowledge Mining System, BioKMS, which integrates AI method in order to solve the pressing needs of biomedical researchers and space science scientists, ultimately expecting to use AI to mine domain knowledge, inspire scientific research ideas, and thus drive scientific research. With user-friendly interactive interface and diversified functions, BioKMS fully takes into account the needs and expectations of researchers in real scientific research scenes. In the traditional biomedical field, BioKMS is a novel knowledge mining service platform that combines knowledge graph and deep learning model. To add to the novelty of the system, BioKMS innovatively couples knowledge graph in the space life science to realize deep mining of knowledge. With the continuous enrichment of data and the continuous optimization of AI model, it is our conviction that BioKMS can shed substantial light on the future scientific research both on the ground and in space.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData sharing is not applicable to this article as no datasets were generated or analysed during the current study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors' contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll Authors have made significant contributions to the work. YZ. Wen: conceptualization, investigation, data curation, methodology, supervision, writing-original draft, review and editing. YX. Han: conceptualization, investigation, data curation, methodology, supervision, writing-original draft, review and editing. JF. Sun: conceptualization, investigation, data curation, methodology, supervision, writing-original draft, review and editing. All authors reviewed the manuscript and approved the final version.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors' information\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eYZ. Wen (Email: [email protected]) is a graduate student in the School of Biological Sciences and Medical Engineering of Southeast University and the Jiangsu Key Laboratory for Biomaterials and Devices.\u003c/p\u003e\n\u003cp\u003eDr. YX. Han (Email: [email protected]) is a faculty advisor to graduate student in the School of Biological Sciences and Medical Engineering of Southeast University and the Jiangsu Key Laboratory for Biomaterials and Devices.\u003c/p\u003e\n\u003cp\u003eProf. Dr. JF. Sun (Email: [email protected]) is a Director in the School of Biological Sciences and Medical Engineering of Southeast University and the Jiangsu Key Laboratory for Biomaterials and Devices.\u003c/p\u003e\n\u003ch2\u003eAvailability and requirements\u003c/h2\u003e\n\u003cp\u003eProject name: BioKMS\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eProject home page: http://8.130.71.195:5000/\u003c/p\u003e\n\u003cp\u003eOperating system(s): Windows 10/11\u003c/p\u003e\n\u003cp\u003eProgramming language: Python\u003c/p\u003e\n\u003cp\u003eOther requirements: Python 3.10.9 or higher, Java 17.0.4 or higher, Neo4j Desktop 1.5.8 etc.\u003c/p\u003e\n\u003cp\u003eLicense: GNU GPL\u003c/p\u003e\n\u003cp\u003eAny restrictions to use by non-academics: licence needed\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eIslam MT, Xing L. A data-driven dimensionality-reduction algorithm for the exploration of patterns in biomedical data. NAT BIOMED ENG. 2021;5(6):624-35.\u003c/li\u003e\n\u003cli\u003eZhao S, Su C, Lu Z, Wang F. Recent advances in biomedical literature mining. BRIEF BIOINFORM. 2021;22(3):a57.\u003c/li\u003e\n\u003cli\u003eFensel D, Şimşek U, Angele K, Huaman E, K\u0026auml;rle E, Panasiuk O, et al. Introduction: what is a knowledge graph? Knowledge graphs: Methodology, tools and selected use cases. 2020:1-10.\u003c/li\u003e\n\u003cli\u003eChen X, Jia S, Xiang Y. A review: Knowledge reasoning over knowledge graph. EXPERT SYST APPL. 2020;141:112948.\u003c/li\u003e\n\u003cli\u003eDong S, Wang P, Abbas K. A survey on deep learning and its applications. COMPUT SCI REV. 2021;40:100379.\u003c/li\u003e\n\u003cli\u003eZhang X, Liang L, Liu L, Tang M. Graph neural networks and their current applications in bioinformatics. FRONT GENET. 2021;12:690049.\u003c/li\u003e\n\u003cli\u003eZhang M. Graph neural networks: link prediction. Graph Neural Networks: Foundations, Frontiers, and Applications. 2022:195-223.\u003c/li\u003e\n\u003cli\u003eMohamed SK, Nov\u0026aacute;ček V, Nounu A. Discovering protein drug targets using knowledge graph embeddings. BIOINFORMATICS. 2020;36(2):603-10.\u003c/li\u003e\n\u003cli\u003eAlshahrani M, Hoehndorf R. Semantic disease gene embeddings (smudge): phenotype-based disease gene prioritization without phenotypes. BIOINFORMATICS. 2018;34(17):i901-7.\u003c/li\u003e\n\u003cli\u003eDimitrakopoulos C, Hindupur SK, H\u0026auml;fliger L, Behr J, Montazeri H, Hall MN, et al. Network-based integration of multi-omics data for prioritizing cancer genes. BIOINFORMATICS. 2018;34(14):2441-8.\u003c/li\u003e\n\u003cli\u003eNita G, Georgoulis M, Kitiashvili I, Sadykov V, Camporeale E, Kosovichev A, et al. Machine learning in heliophysics and space weather forecasting: a white paper of findings and recommendations. arXiv preprint arXiv:2006.12224. 2020.\u003c/li\u003e\n\u003cli\u003eWaisberg E, Ong J, Zaman N, Kamran SA, Lee AG, Tavakkoli A. A non-invasive approach to monitor anemia during long-duration spaceflight with retinal fundus images and deep learning. LIFE SCI SPACE RES. 2022;33:69-71.\u003c/li\u003e\n\u003cli\u003eKilicoglu H, Shin D, Fiszman M, Rosemblat G, Rindflesch TC. SemMedDB: a PubMed-scale repository of biomedical semantic predications. BIOINFORMATICS. 2012;28(23):3158-60.\u003c/li\u003e\n\u003cli\u003eArnold P, Rahm E. SemRep: A repository for semantic mapping. 2015.\u003c/li\u003e\n\u003cli\u003eLee J, Yoon W, Kim S, Kim D, Kim S, So CH, et al. BioBERT: a pre-trained biomedical language representation model for biomedical text mining. BIOINFORMATICS. 2020;36(4):1234-40.\u003c/li\u003e\n\u003cli\u003eDouze M, Guzhva A, Deng C, Johnson J, Szilvasy G, Mazar\u0026eacute; P, et al. The faiss library. arXiv preprint arXiv:2401.08281. 2024.\u003c/li\u003e\n\u003cli\u003eHamilton W, Ying Z, Leskovec J. Inductive representation learning on large graphs. Advances in neural information processing systems. 2017;30.\u003c/li\u003e\n\u003cli\u003eRachner TD, Khosla S, Hofbauer LC. Osteoporosis: now and the future. The Lancet. 2011;377(9773):1276-87.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"biomedical knowledge graph, space life science, link prediction, software system","lastPublishedDoi":"10.21203/rs.3.rs-5232213/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5232213/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground:\u003c/strong\u003e In recent years, Artificial Intelligence (AI) technology, represented by deep learning, has flourished, widely empowering various scientific research. At present, it is worth-noting that biological science and computer technology are developing rapidly and massive biomedical data are accumulating, accompanied by the development of advanced AI algorithms and nascent AI-enabled biomedical research opportunities. Researchers learn to combine AI with real biomedical scenarios, expecting to improve the efficacy of research. However, it is a daunting task for most biomedical researchers, who might lack the knowledge of computer science, to use AI to assist work. Therefore, an open tool integrating advanced AI methods is needed to liberate biomedical researchers from “daily chores” and assist scientific research progress. It is desirable that researchers with little knowledge of AI may still make good use of AI technology to explore scientific clues and promote scientific research.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults:\u003c/strong\u003e Here, armed with the AI technology, we develop Biomedical Knowledge Mining System (BioKMS), a web-based biomedical knowledge mining system encompassing both biomedical and space science knowledge, which demonstrates the knowledge graph and link prediction model. More specifically, the knowledge mining refers to the use of effective methods to mine the internal relationships contained in a large number of data, providing scientific clues for biomedical studies. On the one hand, BioKMS enables researchers to explore the links between domain entity and obtain the visual presentation of the knowledge graph, intuitively capturing the domain-related information; On the other hand, researchers can carry out the link prediction with the encapsulated SOTA deep learning model, which predict the potential association, further revealing potential research objects and assisting decision-making by AI. In addition, in order to improve the timeliness and stability of the data, BioKMS provides flexible data feedback interface that researchers can submit new data, thus cementing the optimization of the subsequent algorithm model.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions:\u003c/strong\u003e In this study, we proposed a novel system to help biomedical scientists use AI to predict and discover potential biological mechanisms. In addition to serving the traditional ground biomedical research, the system integrates the knowledge graph and deep learning model of space life science field to assist the space biomedical research. The BioKMS can be accessed via this URL: http://8.130.71.195:5000/.\u003c/p\u003e","manuscriptTitle":"BioKMS: A Biomedical and Space Science Knowledge Mining System Based on Knowledge Graph and Link Prediction Model","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-11-19 14:26:44","doi":"10.21203/rs.3.rs-5232213/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"91b7ce59-c2c5-4d39-a878-8f816993b687","owner":[],"postedDate":"November 19th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-11-19T14:26:44+00:00","versionOfRecord":[],"versionCreatedAt":"2024-11-19 14:26:44","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5232213","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5232213","identity":"rs-5232213","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}