MARS: An Open-Source Application for Managing and Searching Scientific Metadata

When conducting scientific experiments, vast collections of data and subsequently scientific metadata are generated. Maintaining the set of FAIR (findability, accessibility, interoperability, and reusability) principles is often challenging given large datasets and various metadata standards [1]. MARS (Metadata Aggregator for Reproducible Science) aims to address this challenge as an open-source application by providing a structured interface on top of database storage for the management and querying of scientific metadata, particularly in the biological sciences. MARS introduces a nomenclature for a metadata abstraction that is intended to be customizable and scalable across scientific contexts. A core architectural feature of MARS is the ability to granularize units of metadata into single “entities”. Entities may represent a physical or digital object, ranging from whole animals to brain slice images from that same animal. Raw data are not stored in MARS, rather external links to the relevant data storage platforms are stored within entities. Graph-like connections can be used to establish relationships between entities, and a “project” grouping system within MARS allows entities to be grouped with each other to represent contexts such as experimental output, physical storage contents, or data presented in a publication. To specify metadata associated with an entity, MARS uses key-value pairings known as “attributes”. Attributes can be used to express metadata on a per-entity basis, or attribute templates can be created and reused across multiple entities. Attributes are named uniquely, and the values are restricted to types including “date”, “string”, “number”, or a set of fixed options defined in a drop-down menu. MARS also provides several features useful for scientists, such as an ORCID ID authentication requirement for all users, metadata import from CSV, Excel, and JSON files, multiple metadata export options, and the ability to construct and execute queries across all stored metadata. MARS uses a React frontend written in TypeScript and utilizes components from the Chakra UI library. The Node.js backend couples Express.js with GraphQL to exchange data with a MongoDB database, and is currently hosted on Microsoft Azure. MARS has been deployed successfully within a neuroscience laboratory, managing hundreds of unique entities and 6 active users. Intuitive training for new users is incorporated within the software for each step required to manage entities and attributes across the platform. Potential future deployment targets for MARS include other laboratories and core facilities at Washington University in St. Louis.