Disbiome database: linking the microbiome to disease

For many years, it has been believed that the human body has a microbial cell content which exceeds the total amount of human somatic cells by tenfold [1]. More recently, it has been estimated that this ratio between microbial and human cells is closer to 1:1 [2]. The collection of these microorganisms is termed ‘microbiota’ and the collective genomes of all the microorganisms of these microbiota are defined as the microbiome [3]. The main part of this microbiota is situated in the gut, in which the numbers and complexity increases from the stomach to the colon [4, 5]. Other anatomical sites which have their own microbiome are the lungs, skin, vagina, eyes, placenta, ear, oral cavity and sino-nasal compartment. The composition of the microbiome varies by anatomical site (e.g. between the gut and skin), between individuals and even over time [6, 7]. The microbiome composition can change due to factors such as dietary changes including pre- and probiotic use, antibiotic and other medicine use, age or disease and is moreover dynamic on its own [8, 9] . The microbiota composition and its correlation with health/disease is thus considered a multifactorial process. An active lifestyle can influence the gut microbiota composition, enhancing diversity and promoting bacterial communities associated with healthy individuals, which tend to be dominated by species such as Faecalibacterium prausnitzii, Roseburia hominis and Akkermansia muciniphila [10, 11]. Moreover, the mode of delivery has a major influence on the microbiome composition of the new born. After vaginal delivery, the baby’s microbiome resembles the mother’s genital and gastrointestinal tract while bacteria of the skin appear to be more abundant after caesarian section [12]. Several intrinsic and extrinsic factors influence the development and variation of bacteria in infants. Genetics and epigenetics, environmental factors like geography and diet (breastmilk or formula fed) all affect development of the microbial population [13, 14]. However, a lot of questions concerning the development of the fetus and neonate microbiome are still open [15].

While it was initially thought that microbes are mainly commensals whose only benefit is controlling the population of pathogenic bacteria, there is compelling evidence that the gut microbiome also has health-influencing effects, playing roles in i.a. digestion, inflammation, intestinal integrity and development of the immune system [16]. Production of microbial metabolites are a key driver in these processes. Host health thus appears to be closely related to a homeostatic balanced relationship with the microbial inhabitants. Several diseases are associated with an altered microbiota composition, such as obesity [17], diabetes [18], Crohn’s disease [19], ulcerative colitis [20], autism [21], bacterial vaginosis [22] and psoriasis [23]. These alterations are not limited to the location of the disease. As an example, alterations of the gut microbiota are seen in a variety of central nervous disorders, supporting the presence of a gut-brain axis [24, 25]. At this point, it remains to be elucidated whether the observed microbiota differences in various disease states are a symptom of the disease or have a more causal effect [6]. Suppressing clinical dysbiosis and restoring the altered microbiome to a ‘healthy’ microbiome can be a potential approach to improve host health. Potential therapeutic options include narrow spectrum antibiotics, probiotics, prebiotics, dietary interventions and fecal transplantation [16].

Different microbiology databases for research are available. There are databases covering different microbial subjects such as genomic resources (e.g. IMG) [26], protein families (e.g. Pfam) [27], diversity (e.g. SILVA) [28], model organisms (e.g. EcoCyc) [29], pathogenesis (e.g. EuPathDB) [30], transport and metabolism (e.g. TCDB) [31] and signal transduction and gene regulation (e.g. MiST) [32]. However, a database covering microbiome differences in different disease states is, to our knowledge, currently missing. Seen the exploding data of microbiome alterations in different disease states, we present the Disbiome database, collecting and organizing this information (https://disbiome.ugent.be). Disbiome encompasses microbiome differences between patients and controls together with the used detection method and sample type. This database differs from other comparative tools such as MG-RAST as it presents comparisons between patient and control data in a clear and concise manner to the broader audience in a programmatically accessible way using the JSON export format [33]. Disbiome can be valuable for every researcher in the field of microbiology to rapidly and easily find bacterial species possibly correlated to specific diseases to further explore its mode of actions and interaction mechanisms with the host. It can speed up translational research in microbiome modulations (by either probiotics, prebiotics and microbiota transplantation) for treating a variety of diseases. In addition, it can serve as a new disease classification system based on microbiome changes. Currently, the database includes over 190 different diseases and 800 different organisms. Changes in organisms are detected by over 25 different detection methods (e.g. qPCR, next-generation sequencing,…) in 50 different sample types (e.g. faeces, skin swabs, tissue biopsies,…).

To list all relevant data, a relational database was constructed [34]. A relational database separates the design of the data from its physical representation. Data are designed as tables where the rows represent distinct entities and the columns represent various attributes of those entities [35]. The schematic database model is given in Fig. 1. This visual representation represents the structure of the database. A block represents an entity type, where the table’s columns represent the entity type’s different attributes. Such entity type can have an unlimited number of entities. In the physical database, each entity type translates to a table where each attribute represents a column of that table and each entity is represented by a row. The central entity in the Disbiome database is ‘Experiment’, representing the microbiome difference between a patient and control. Every experiment has a qualitative outcome (elevated or reduced) and is related to different parameters. The experiment is linked to the appropriate publication (Publication ID), showing a microbial (Organism ID) difference between a sample (Sample ID) of a patient (Disease ID, Host ID) and a control subject (Control ID) using a specific detection method (Method ID). This sample originates from a specific location (Location ID). The microbial difference can be presented by absolute quantities between the patient and control or by a ratio. In the case of absolute data, a specific response unit, dependent on the used detection method, is present (Response ID). When absolute data is not available, the microbial differences are only presented by the qualitative outcome. Every host and publication is linked to different host and publication parameters respectively and diseases are linked to their classification in the Medical dictionary for Regulatory Activities (MedDRA). The storage of Disbiome was implemented using PostgreSQL, an open source database. This is accessed by LimeDS, a framework developed at Ghent University, providing a web service with which the website (disbiome.ugent.be) communicates [36]. Several search options are present in the Disbiome database. Organisms, diseases and detection methods can be used as queries and will give an overview of the experiments related to this organism, disease or detection method. From this overview page, detailed information about the experiment can be obtained.

Literature data was collected by using the search engine PubMed, covering the period 2009–2018. The search queries [(‘microbiota’ OR ‘microbiome) AND (‘health’ OR ‘disease’)] and [microbiome alterations] were used. This search gave approximately 20.000 publications. Based on title (exclusion of duplicates, reviews, animal studies, studies on effect of medication, symposium/meeting abstracts as well as non-English language papers), only around 1.000 publications were withheld and based on the abstract (only case-control studies) around 500 publications were found to be suitable for insertion in the database. The obtained literature was processed manually and all relevant information was put in the database. Currently only human data is incorporated but information of other species (e.g. mice, rats,…) will be included as well. The database will be updated manually every 3 months. An automated updating system is being developed for further versions of the database. Additionally, authors of publications can inform us of missing manuscripts by using the ‘Submission’ link.

Our main objective was to construct a database giving a clear and rapid overview of all bacterial species which are differentially present in a particular disease.

Literature data on microbiome alterations in different disease states is vastly increasing. Disbiome (https://disbiome.ugent.be/) provides an organized overview of this rapidly expanding field of knowledge. Together with the used sample, detection method, methodological details and host information, quantitative data of micro-organisms in patients and controls from a specific experiment are presented. In addition, different reporting parameters of the concerned publications are presented. This is the first database giving a relation between the health status of the host and its microbiota composition.

Project name: Disbiome.

Project home page: https://disbiome.ugent.be/.

Browser: Google Chrome, Microsoft Internet Explorer, Mozilla Firefox.

Lisence: none.

Restrictions for non-academic users: none.