CROssBAR: Comprehensive Resource of Biomedical Relations with Deep Learning Applications and Knowledge Graph Representations

Abstract

Systemic analysis of available large-scale biological and biomedical data is critical for developing novel and effective treatment approaches against both complex and infectious diseases. Owing to the fact that different sections of the biomedical data is produced by different organizations/institutions using various types of technologies, the data are scattered across individual computational resources, without any explicit relations/connections to each other, which greatly hinders the comprehensive multi-omics-based analysis of data. We aimed to address this issue by constructing a new biological and biomedical data resource, CROssBAR, a comprehensive system that integrates large-scale biomedical data from various resources and store them in a new NoSQL database, enrich these data with deep-learning-based prediction of relations between numerous biomedical entities, rigorously analyse the enriched data to obtain biologically meaningful modules and display them to users via easy-to-interpret, interactive and heterogenous knowledge graph (KG) representations within an open access, user-friendly and online web-service at https://crossbar.kansil.org. As a use-case study, we constructed CROssBAR COVID-19 KGs (available at: https://crossbar.kansil.org/covid_main.php) that incorporate relevant virus and host genes/proteins, interactions, pathways, phenotypes and other diseases, as well as known and completely new predicted drugs/compounds. Our COVID-19 graphs can be utilized for a systems-level evaluation of relevant virus-host protein interactions, mechanisms, phenotypic implications and potential interventions.

RTX-KG2: a system for building a semantically standardized knowledge graph for translational biomedicine

“…The Hetionet project [41] developed concept types and relationship types specifically for knowledge representation for drug repurposing; these types were expanded in the Scalable Precision medicine Knowledge Engine (SPOKE) database [42, 43]. CROssBAR-DB [44] keeps its source datasets separate from one another and provides an interface for constructing a query-specific, integrated knowledge graph. The Reasoning Over Biomedical Objects linked in Knowledge Oriented Pathways (ROBOKOP) graph [45, 46] uses concept and relationship types from the Biolink metamodel.…”

“…While both approaches have their strengths, from our work on the predecessor RTX-KG1 system and from the present work, we found that an ETL approach has significant advantages in terms of scalability, reproducibility, and reliability. In terms of automation frameworks, previous efforts have used general-purpose scripting languages [34, 36, 41, 42, 45, 49, 51], batch frameworks [44], declarative rule-based build frameworks [31, 33, 52], and parallel build systems such as Snakemake [53] (EpiGraphDB). Liu et al reported [52] choosing the Snakemake [53] build framework specifically because of its high performance (i.e., parallel capabilities).…”

“…For persistence, knowledge systems have used relational databases [30], distributed graph databases [33], multimodal NoSQL databases [31], RDF triple-stores [34, 36], document databases [35], document-oriented databases [44, 50], and—with increasing frequency [35, 41, 42, 45, 47, 49]—the open-source graph database Neo4j (github:neo4j/neo4j). Knowledge systems have also differed in terms of the ingestion method used in their construction; many systems [31, 34–36, 44] utilized an extract-transform-load (ETL) approach, whereas others [45, 49, 50] used API endpoints to query upstream knowledge sources; one [47] blended both ETL and API approaches for knowledge graph construction. While both approaches have their strengths, from our work on the predecessor RTX-KG1 system and from the present work, we found that an ETL approach has significant advantages in terms of scalability, reproducibility, and reliability.…”

“…Previous efforts to develop integrated biomedical knowledge systems have used various database types, architectural patterns, and automation frameworks. For persistence, knowledge systems have used relational databases [34], distributed graph databases [33, 57], multimodal NoSQL databases [35, 57], RDF triple-stores [41, 42, 58], document-oriented databases [32, 46, 54], and—with increasing frequency [36, 37, 39, 44, 54]—the open-source graph database Neo4j (github:neo4j/neo4j). Knowledge systems have also differed in terms of the ingestion method used in their construction; many systems [32, 35, 41, 42, 54] utilized an extract-transform-load (ETL) approach, whereas others [44, 46, 59] used API endpoints to query upstream knowledge sources; one [39] blended both ETL and API approaches for knowledge graph construction.…”

“…At the same time, advances in natural language processing (NLP) [19][20][21][22][23][24] have enabled systematic extraction of structured knowledge from the biomedical literature, such as the Semantic MEDLINE [25] Database (SemMedDB) [26]. Community-driven ontology development [27][28][29][30][31], literature curation, and the use of NLP together have driven growth of structured biomedical knowledge-bases, albeit in forms that are not semantically interoperable due to the use of different systems of concept identifiers, semantic types, and relationship types [32].…”