gms | German Medical Science

Introduction: In order to understand a model molecular mechanisms in diseases and support the development of novel therapeutics, disease maps are being developed to represent existing knowledge on disease mechanisms in a computationally readable and comprehensive manner. To make them publicly available, the open-source platform MINERVA was developed by Gawron et al. in 2016 [1]. Disease maps can i.a. be used to elucidate the mechanism of action of promising drug candidates from high-throughput screens, identify side effects and adversary reactions. This project aims to create a MINERVA plugin to link disease maps and application specific databases of drug candidates.

State of the Art: The MINERVA Platform comes with its own drug and chemical search. The user can search for drugs or chemicals in different databases for known targets in the map. Furthermore, the MINERVA Drug reactions plugin aims at exploring adverse reactions of drugs which are interacting with entities in a given disease map. It connects to a drug-reactions data source and uses MINERVA’s drug search to find the targets of any of the drugs in the database map.

Concept: We previously developed the generic IT-solution CandActBase [2] for the collection and organization of literature data on drug candidates. Our MINERVA-CandActBase plugin links data encoded as disease maps in the MINERVA platform and these application-specific databases of potential drug candidates. We extended the CandActBase database with data on compound-gene interactions from ChEMBL database [3] and the Comparative Toxicogenomics Database (CTD) [4]. Our plugin offers three main functionalities, the show all function, the compound search and the target search.

Implementation: The Code-structure of the plugin is based on the existing MINERVA plugin starter-kit. Interactions between chemicals and genes were extracted for every compound from the CandActBase from ChEMBL and CTD. The data is stored in JSON format on the hosting server to ensure fast loading times. One of the two main JSON files contains the compound identifier used in CandActBase and the name of the genes targeted by each compound. Vice versa, the second JSON file includes every gene or protein name from the graph and a list of compound IDs that target the given gene or protein. The plugin will first be available via the CandActBase Website, but a standalone version is planned, which can be downloaded and used in any MINERVA instance.

Lessons Learned: Our plugin is easily adaptive to more and different databases, if their data can be transformed to a common data format. This highlights one of the biggest hurdles we had to overcome in this project, which is due to the huge variety in data structure of different databases. There is no standardized data format to rely on. Even basic information, like common and unique identifiers for chemicals, differ across databases. This is naturally due to the variable use cases of databases, but it would simplify the development of cross-database applications considerably.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.