Many studies aimed to construct an automated gene expression analysis platform for researchers. However, they lack an integrated data model for analyzing heterogeneous data. In order to address this issue, we created a biological data integration platform for transcriptome analysis (BiDIP) for managing various kinds of databases. As part of this platform, we developed a biological interaction data model (BIM). Also we provide a Web application and OpenAPIs that allow users to search for connections among multiple databases conformant to the proposed data model. In this paper, we will present the current status of the platform as well as its future research venues.
Gene expression analysis has been a valuable research venue that provides for explanatory power to specific biological phenomena, given a biological context for a patient’s histology or medicinal intervention. With the advent of microarray technology, transcriptome analysis has greatly overcome the limitations resident in studying individual genes, which offers fragmented and insufficient explanation for biological phenomena. Microarray technology enables us to measure and identify compositions of transcriptome under certain biological conditions. It, however, poses a significant challenge to biologists, especially those who are not well-versed in using the technology. Seemingly small, but still challenging to biologists, was the necessity for adequate amount of knowledge in coding required for data processing and statistical analysis of the data gained from microarray experiments. Almost 10 years after the introduction of the technology, we have acquired standard protocols for microarray analysis, though biologists still quite often need helping hands of computer experts or of companies that provide experiment tool kits.
We aimed to develop a platform that facilitates transcriptome analysis called Online Transcriptome Analysis Pipeline (OnTAP). Two major components constitute OnTAP: a Biological Data Integration Platform (BiDIP) and analysis modules including data pre-processing. There are four sets of analysis modules: gene set enrichment analysis, gene network analysis, drug target prioritization analysis, and miRNA target prioritization analysis.
BiDIP consists of four main components: 1) a comprehensive database model, BIM (Biological Interaction data Model) that encompasses 4 types of biological databases; 2) 4 integrated databases, which deal with PPI (Protein-Protein Interaction) databases, DGI (Drug-Gene Interaction) databases, microRNA databases, and pathway databases; 3) BiDIP browser, and 4) OpenAPI. BiDIP provides a unified view on various biological databases facilitating and streamlining transcriptome analysis, alleviating some burden off biology researchers. This paper presents OnTAP in general, and BiDIP in particular. 2
Studies abound that aimed to provide streamlined services for microarray analysis:
ArrayPipe [
The notable drawbacks of the previous studies are: 1) limited coverage/volume of data - it is arguably true that the higher volume of, and wider coverage of, data will plug the gaps in our knowledge for biological phenomena; 2) absence of a comprehensive and inclusive data model by which to view and analyze heterogeneous databases. 3
This section presents detailed explanation on BiDIP, with focus on Biological Interaction data Model (BIM), integrated databases, Web-based browser, and Open API. 3.1
The core relations that BIM is designed to express are gene-gene, gene-drug, genemiRNA relations, and pathway data. As shown in Fig. 2, the top classes in BIM are process, interaction and biochemical entity. Biochemical entity represents objects involved in a biological interaction. An interaction connects two biochemical entities. A sequence of interactions makes a process. In a process, the order of the interactions carries significant meaning. Fig. 3 shows a more detailed overview of BIM. Biochemical entity contains two main subclasses: gene and drug. Gene is defined as a biological region with a specific function. Gene has three subclasses: 1) DNA-specific region that will become a protein or functional RNA, 2) transcribed RNA from DNAspecific region, 3) translated protein from transcribed RNA. RNA has two subclasses: miRNA and mRNA. Drug represents any material used to treat or prevent diseases. In transcriptome analysis, the distinction between gene and its subclasses has not been used in general. However, we have made that distinction to reflect biological reality, which allows BIM to link with other ontologies that require such distinction. Biochemical entity is linked to EntityFeatures class with properties expressing full-name, genomic region, function annotation, and molecular weight.
Interaction has subclasses representing various unions of biochemical entities such as Gene to Gene Interaction (GGI) and Drug to Gene Interaction (DGI). GGI has miRNA to Gene Interaction (MGI) subclass, and Protein-Protein Interaction (PPI) subclass. GGI has two gene participants, MGI has mRNA and miRNA participants, PPI has two protein participants and DGI has gene and drug participants. Interaction also has a link to Publication class, each object of which is identified by PubMed id (PMID). Process has a subclass pathway representing a sequence of GGI. 3.4
A process has a 1: N relationship with its interactions and each interaction has 2 biochemical entities using a participant property. Participant property has two sub properties: participant gene and participant drug. participant gene has three sub properties: participant_mRNA, participant_miRNA, and participant_ptn. Each interaction has a datasource property to represent the origin of database. The xref (external reference) property is used in two cases 1) to link publication objects to publication url 2) to link a biochemical object to an external site for more information. To express a sequence in a pathway we used the nextstep property, as shown in Fig. 5. 3.5
Biological databases are integrated based on BIM as shown in Table 1. We extracted
human species (Homo sapiens) records from the databases and then converted Gene ID
to official Gene Symbol by the HGNC standard.
KEGG, DGIdb [
This section presents an example of how a gene is expressed in BIM. An ion channel gene TRPC1 is used as an example in Fig. 4. TRPC1 is linked to Protein STIM1 with PPI (PPI 307419), to miR-124-3p with MGI (MGI 235456), and to Dantrolene with DGI (DGI 51715). Each interaction instance has a datasource property and an xref property. Fig. 5 shows a serotonergic synapse pathway that includes TRPC1. A pathway instance has more than one GGI component and a GGI instance has 2 participating genes. The order of each GGI instance is determined by using the nextstep property to represent the pathway flow. A Web-based application, BiDIP Browser, and open-access APIs were developed. Both services were designed and implemented to enable efficient retrieval of gene-centric interaction information on drugs, miRNAs, pathways and proteins based on BIM. As shown in Fig. 6, the browser has 5 panels: 1) Finding genes & Gene browsing history, 2) drug list, 3) miRNA list, 4) gene (on protein layer) list, and 5) pathway list. Given a keyword entered in the Finding Genes box, a list of candidate genes is listed whose symbols start with the keyword. Once a gene is selected from the search results, it is first recorded and listed as a previously-selected gene in the Gene browsing history list. The drug, miRNA, gene and pathway panels show elements interacting with the selected gene. Clicking the ‘Save browsing history’ button will enable the user to save his/her browsing history as an XML file. The browser is available at http://147.47.41.58:8080/BiDIP_Browser/. Most of the information the user can access from the browser can be obtained by using our RESTful web services, BiDIP OpenAPI; and third parties can use the services to develop their own systems and applications that make use of the BIM-based gene-centric interaction datasets. Table 2 gives a list of the web services with examples.
Fig. 6. Screenshot of BiDIP Browser
In this paper, a biological data integration platform, BiDIP, is presented. BiDIP integrates four sets of interaction databases into a common data model. The differentiating factors of this study that distinguish it from the extant similar gene expression analysis studies are as follows: 1) development of comprehensive data model, BIM, for interaction databases; 2) creation of consolidated databases for four types of heterogeneous interaction databases; 3) two points of access, BiDIP Browser and open APIs, to the database. BiDIP enables researchers to gain access to integrative biological information about genes of interest.
The focus of this study has been to develop a solid basis for interaction-focused databases. Hence, there are certain details left out for now such as epigenetic information, DNA variant, and etc. These data will be progressively added to BIM.
Finally, BiDIP is part of a larger project called OnTAP. We are currently in the process of developing data input modules, data pre-processing modules, and data analysis modules for full deployment of OnTAP for general use.
This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT and future Planning (NRF-2014R1A2A1A11049728)