Reproducibility for scientific analysis through formal data-driven software validation - a pilot infrastructure for neurophysiology

• Applicants: Dr. Michael Denker; Professor Dr. Thomas Wachtler-Kulla
• Subject Area: Experimental and Theoretical Network Neuroscience
• Term: since 2026
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 577209274

Reproducibility in research requires not only that results of scientific analyses can be recreated using the exact same procedure, but also that results do not depend on the specific method used. With respect to analysis research software, this requirement implies that different implementations of the same method should yield the equivalent output, and that they should be validated to ensure compatibility. In this project we propose to set up an infrastructure that enables researchers to validate their code against reference datasets and compare it to other implementations. To achieve a sustainable and general solution, we will define a data model for analysis results and an ontology for validation, which will be used to set up a service that can be seamlessly integrated in researchers' code development workflows and can be extended by contributions from the scientific community. For efficient development we will focus on use cases from the field of systems neurophysiology, an area where, on one hand, a large amount of custom code is written in individual laboratories and, on the other hand, resources exist that can serve as building blocks for the infrastructure. Nevertheless, the solution will be designed to be easily adaptable to other fields. To our knowledge this is the first systematic approach to foster the practice of validation in research software engineering. Providing easy access to code validation, it will have the potential to enhance the reliability, credibility and reproducibility of analysis research software.

• DFG Programme: Research data and software (Scientific Library Services and Information Systems)