Computable model reuse improves credibility?

With the drive to translate computational models into clinical applications, the perceived credibility of models is a hugely significant factor. Unambiguous and detailed model provenance is crucial evidence in the evaluation of model credibility. Traceability of model reuse in a standardised, computable, and precise manner eases the collection of such evidence.

Modelling standards such as CellML and SBML have existed for over 20 years. While they have undoubtedly had a significant impact on reproducibility, their impact on model reuse is less clear. There are three main goals of this project. The first is to obtain a detailed understanding of the current state of model reuse in the systems biology community to derive insight and develop forward-looking proposals to improve model reuse. The second is to leverage these insights in combination with the recent emergence of modelling modularity practices and tooling to develop novel modelling methods, tools, and/or guidelines to improve "computable" model reuse. In achieving these goals, we will be well-positioned to address the third goal by answering the question posed in this project's title.

Anecdotal evidence collected in the curation of the hundreds of models available in public repositories suggests that while publications of new models often cite any existing models from which they derive, the actual "model reuse" is implemented via copy-and-paste of earlier implementations or a complete re-implementation from the published literature. In reusing models in this manner, it is difficult to infer anything about the credibility of the new model from the original models. With a "computable" model reuse approach, however, there are standardised and persistent links between models, unambiguous descriptions of any modifications, and a range of semantic goodies that can be used to enhance the credibility of new models from the credibility of the original models.

Page expires: 1 May 2024