Data and Model Sharing Group

Science is a social activity requiring transparency, collaboration, and critical evaluation of published research. The bulk of research today is still communicated via peer-reviewed publications, and journals represent the primary gateway by which research is disseminated. One of the core aspects of the scientific method is the need to reproduce results. This is one of the primary distinguishing features that makes the scientific method so successful.

However, there is today significant pressure by research insitutions for researches to publish in volume and more rapidly. One of the consequences of this is that the reproducibility of published research becomes secondary. The lack of reproducibiltiy diminishes the stature of science as a reliable methodology and can affect policymakers in government and industry and instill a lack of trust by the general population. One approach to remedy this situation is to encourage journals to make sure that published work is reproducible. This means developing policies and a more open culture that encourages sharing of empirical data, and models alongside the publication. The data and model sharing group therefore enourages members of the IMAG community and beyond to make every effort to ensure their models, data and computations are publcally avaialsble and reproducible. There are three simple rules that can greatly improve the situation, they include:

1. Stating the software used in the study, including the particular version used.
2. Providing machine readable code in supplements or uploaded to established repositories,
3. Asking a third-party to test that your methods section is free from error and of sufficient detail to reproduce the 
results presented in the paper.  

The last rule can be surprisingly effective and often eliminates many of the most obvious issues when sharing data, models and code.

Data- and model-sharing

To make it easier to conduct reproducible biochemical modeling, the Center for Reproducible Biomedical Modeling is developing tools that simplify model building, annotation, simulation, and visualization. However, if the model, results, and documentation associated with published modeling studies are not accessible, and the modeling workflow is not transparent, reproducing the model and its simulation results remains difficult and reuse is impossible.

Therefore, we recommend that all model artifacts produced during the modeling workflow be publicly shared to facilitate reproducibility and reuse, as emphasized by the Findability, Accessibility, Interoperability, and Reusability (FAIR) principles. Following the FAIR principles will ensure that models can be downloaded and manipulated by independent research groups, allowing published results to be validated, and will allow complex models to be built by integrating previously-published work on components of the system. We encourage modelers to disseminate packages of artifacts alongside publications with an open-source license and to deposit these packages in version-controlled public repositories.

Alignment with the NIH Strategic Plan for Data Science