You are here

Standards, Standards, Standards...

Maciej Swat's picture

DDMoRe is all about interoperability of models, input data and results. Currently, the M&S process remains fragmented by the lack of common tools and the existence of many software packages, which implement proprietary formats for representing models and data, thus, limiting the potential reuse of models and analysis methods. Indeed, the lack of interoperability across tools results in a large duplication of effort since, for example, performing all the relevant analyses with a single model requires recoding the model and the data each time. This significantly limits the integration and the sharing of existing knowledge as well as the possibility of building on it in an innovative way. DDMoRe intends to address these problems by enabling exchangeability and interoperability of information in the M&S domain. As its core, DDMoRe has standards for the efficient exchange and reuse of models, input and output data across target software and M&S tasks.

Actually, the model should be encoded in a format allowing its translation into an arbitrary target language. To achieve such degree of interoperability a standardisation of a particular domain is required. In this case it means to find the most suitable, tool-agnostic representation for mathematical and statistical models used in Pharmacometrics (PMX) - not quite a trivial task for this complex domain...

Similarly, to enable effective data flow across tasks and information retrieval for post-processing and reporting, one would like to be able to store numerical output in a standardised format.

While developing exchange standards is very challenging, they exist in various computational science areas, some already for more than 15 years. Systems Biology Markup Language (SBML) is one of such examples, an encoding standard for computational models of biological processes (Hucka, 2003), which plays a pivotal function in models dissemination and usage.

PMX lacks behind for a number of reasons. The complexity of this area, characterized by the existence of very heterogeneous target tool languages (e.g. the imperative NMTRAN of NONMEM and declarative MLXTRAN of Monolix), makes its standardisation very demanding. Different notation and vocabulary are yet another burden one has to face when trying to find a common ground in PMX model encoding. On the output side, each tool stores data in different formats and file types, thus, limiting the exchange of information across tasks and tools, and then the modelling capabilities of an M&S workflow.

Over the past five years, DDMoRe has developed two candidates for such exchange standards and both are based on the XML technology (Figure 1). Pharmacometrics Markup Language (PharmML) provides the structure for the encoding of model definition and trial design and modelling steps (Swat, 2015). It is based on the mathematical theory of nonlinear mixed effect models (NLME), representing the backbone of all the target tools used in PMX (Lavielle, 2014). By analysing numerous real-life use cases and literature examples, we have been able to develop and implement a structure capable of encoding models in a tool-independent manner.

Figure 1.

 

As a complementary tool-independent format, the Standard Output (SO) is capable of storing all relevant results produced in the common modelling steps, such as estimation, simulation or optimal design irrespective of the used target software. It is a result of an extensive analysis of results from various target tools.

PharmML and SO, the new kids on the block, cover both model definition and tool output (Figure 2). They have the potential to change the landscape of PMX as we know it today, as they facilitate:

  • Smooth and error-free transmission of models between tools.

  • Use of complex workflows via standardized model and output definitions.

  • Reproducibility of research.

  • Easier reporting and bug tracking.

  • Improved interaction with regulatory agencies with respect to model and data sharing.

  • Reuse of existing model resources, e.g. BioModels database.

  • Development of new tools and methods.
     

Figure 2.

 

Although we have already reached a good coverage of the domain, new applications and model types will certainly pose new challenges on the formats, thus, requiring the continuation of work beyond the duration of the DDMoRe project lifetime. The DDMoRe Foundation will create an ideal environment for the maintenance, development and maturation of our standards.

Maciej J Swat (EMBL - European Bioinformatics Institute, Cambridge, UK) & Nadia Terranova (Merck Institute for Pharmacometrics, Merck Serono, Lausanne, Switzerland)

Further readings:
Hucka, M. et al. (2003) The systems biology markup language (SBML): a medium for representation and exchange of biochemical network models. Bioinformatics 19, 524–531.
Swat, MJ. et al. (2015) Pharmacometrics Markup Language (PharmML): Opening New Perspectives for Model Exchange in Drug Development. CPT Pharmacometrics Syst Pharmacol, 4(6):316-9. 
Lavielle, M. (2014) Mixed Effects Models for the Population Approach Models, Tasks, Methods & Tools. Chapman & Hall/CRC Biostatistics Series