Science Integrity Alliance asks Gillian Currie, postdoctoral researcher at the University of Edinburgh's Institute for Neuroscience and Cardiovascular Research and a member of the iRISE Steering Committee.

iRISE (improving Reproducibility In SciencE) was created in 2023 and works alongside two of the other Horizon Europe projects funded under the same call, “Increasing the Reproducibility of Scientific Results.” What motivated it, and how do the projects avoid duplication?

iRISE and TIER2 both grew out of a meeting in Berlin in January 2022 on the growing problem of irreproducibility in research. We aimed to provide a structured understanding of why irreproducibility happens, along with practical tools and interventions for researchers, funders, publishers and policymakers. iRISE, TIER2 and OSIRIS (also funded under the call) are complementary, approaching the same challenge through different lenses. 

For example, TIER2 and OSIRIS mapped interventions through a scoping review, and iRISE built on that work by developing and validating an automated, AI-assisted approach to create a living evidence base on reproducibility interventions: the iRISE Systematic Online Living Evidence Summary (SOLES). We then formalized our collaboration through the MEta-Research for Research Improvement (MERRI) Collaboration, which is now an open community for anyone working in or around meta-research, not only members of the original consortia.

Explore and Join MERRI

How were your team and stakeholder group chosen, and what has iRISE learned about equity, diversity and inclusion (EDI)?

We deliberately built a multidisciplinary team spanning qualitative and quantitative research, social sciences, statistics, biomedical sciences and law. Junior researchers co-led several work packages, with mentoring from senior colleagues, sharing leadership across different career stages. A mapping exercise guided the choice of a stakeholder group of funders, publishers, researchers, institutes and organizations with broad geographic representation, including people already working on reproducibility.

However, EDI has been a genuine learning experience for us. We first focused on whether interventions might be inequitably accessible or disadvantage particular groups, but we came to understand that EDI must be embedded not only in the outputs but also in the processes, interactions and structures of the research itself. We also wrongly assumed a shared baseline understanding of EDI across the consortium and stakeholder group. In the future, we will establish a shared understanding of EDI from the outset and embed equitable processes throughout the research.

What are iRISE's main outputs, and how far along have you come in your goals?

Our six work packages move from understanding the problem, through identifying and testing solutions, to sharing those solutions with the research community. Outputs so far include:

• a shared glossary of terminology;

• iRISE-SOLES, a publicly accessible overview of candidate interventions across disciplines; 

• best practice guidelines for embedding EDI in reproducibility tools and interventions;

• a report on barriers and facilitators to implementation; and 

• a priority list of interventions integrated into The Embassy of Good Science.

Of our five preregistered intervention studies, two randomized controlled trials and a qualitative study are nearly complete, and a simulation study was recently published. Our Implementation Guide summaries bring together evidence on specific interventions.

Discover iRISE Outputs

Your simulation study evaluating replication success metrics for animal-to-human translation was published in August 2026. Could you briefly describe what you found?

Most drugs that work in animals fail in humans —by some estimates, up to 95% of them. Yet, little research has examined how to measure whether animal findings actually "translate" to humans, with most studies relying on simple significance criteria. This simulation study, led by Rachel Heyard, tested whether replication metrics could assess animal-to-human translation. Nine metrics were evaluated across 648 scenarios using real parameters from a maternal blood pressure meta-analysis. Key takeaways include the observation that no single metric was best: the right choice depends on your goal and data. We recommend that researchers match the metric to the goal and use several metrics together rather than one alone. Metrics became too permissive as heterogeneity rose because small, underpowered animal studies dragged translation power down. But remember, this study covers only the statistical side —safety, pharmacology, and biology still matter.