PharmML: a flexible format for model exchange

For predictive medicine to become a practical reality, scientists must be able to build, share and access accurate, efficient models of disease and drug action.

Pharmacometric modelling is essential for designing cost–effective, reliable clinical trials to test new or repurposed medicines, but until now it has been difficult for researchers to pool their knowledge and develop these models collaboratively between organisations.

PharmML, developed at EMBL-EBI, is helping researchers collaborate on computational models of disease and drug action. An article published in the journal Clinical Pharmacology & Therapeutics: Pharmacometrics and Systems Pharmacology (CPT:PSP) introduces PharmML as a key component of a software platform that improves computational tools for drug discovery, built by the Drug Disease Model Resources consortium (DDMoRe).

“PharmML is an exchange format for pharmacometric and systems-pharmacology models,” explains Lutz Harnisch, Coordinator of the DDMoRe consortium. “The public repository built on PharmML lets researchers in systems biology, quantitative systems pharmacology and pharmacometrics carefully scrutinise a broad range of published models. It provides easy access to models in their context of use, and helps authors enhance their credibility based on the use and reusability of their approaches. These are all vital to improving transparency in scientific communication.”

PharmML is being developed to facilitate smooth, error-free transmission of models between tools, and to make reporting and bug tracking easier. It provides support for the implementation of non-linear, mixed-effect models for the analysis of longitudinal population data from clinical trials in a ‘tool agnostic’ manner.  PharmML also makes it easier to navigate complex workflows and improves interactions with regulatory agencies regarding modelling and simulation.

“As a biostatistician involved in pharmacometrics for several decades, I am very pleased to see such a well constructed and clever language that encompasses all the statistical features needed in this area of computational modelling. It will also be a very valuable tool for teaching Masters-level and PhD students,” says France Mentré, Professor of Biostatistics at Université Paris Diderot and director of an INSERM research team.

“The pharmaceutical industry and regulatory agencies have a long track record of adopting standards because this makes it much easier to achieve real advances in quality, quantity and efficiency,” says Peter Milligan, Head of Pharmacometrics at Pfizer and EFPIA group participant for DDMoRe. “I see the potential for PharmML to be the standard that makes it possible to use a burgeoning variety of quantitative approaches in this sector. Until now, this has been inconceivable simply because of the inherent incompatibility of current modelling tools.”

“The development of PharmML is a brilliant example of what can be achieved when public bioscience research joins forces with pharmaceutical research, and I have no doubt that it will have a significant impact on pharmacometrics,” says Nicolas le Novère, EMBL-EBI alumni and Senior Group Leader at the Babraham Institute. “Standards for pharmacometric modelling developed by the DDMoRe project improve sharing and reproducibility of quantitative studies in drug development. One can use the same mathematical model with different tools and pass the model through different stages of the drug development pipeline – from preclinical to clinical. The result is that these models can be accessible to everyone by way of public model repositories.”

“Designing PharmML has been a fascinating challenge, and not just a technical one,” says Maciej Swat, lead developer of PharmML at EMBL-EBI. “It requires an intimate understanding of both statistical methods used in pharmacometrics, and the needs of the modellers. We’ve been lucky to have so many top European pharmacometric groups and experts in the DDMoRe consortium, sharing their invaluable experience and know-how.”

Leave a Reply

Your email address will not be published.