[Population Modeling] DDMoRE and PharmML/SO introduction
Maciej Swat
mjswat at ebi.ac.uk
Tue Oct 20 03:02:29 PDT 2015
Hi All,
I would like to introduce DDMoRe, an European IMI project, developing an
interoperability platform integrating modelling tools used for the analysis of
clinical data.
Although we don’t develop new estimation methods, I thought it would be
interesting for this list to have some insights into our project.
In one of the work-packages lead by our group, we develop new markup
languages, PharmML/SO for storing and exchange of pharmacometrics
models and associated results. This proofed to be a challenging enough task
given the complexity of pharmacometrics and the variety of models to cover.
DDMoRe project (Drug Disease Model Resources):
The lack of a common exchange format for mathematical models in
Pharmacometrics is similar to the situation ~15 years ago in computational
biology. There, a format called SBML was designed, Systems Biology
Markup Language, which is nowadays widely accepted.
The DDMoRe project is trying to fill the gap by building an interoperability
platform providing support to formulate a model and run it in a target tool.
The key elements of this platform are new model exchange formats.
Pharmacometrics Markup Language (PharmML) provides means
to encode models used in pharmacometrics, i.e. the statistical model,
trial design and modelling steps in ‘tool agnostic’ manner.
The Standardised Output (SO) had been part of PharmML initially, but
is now being developed as a separate format. It is format for flexible
storage of typical results of M&S analyses, enabling effective data flow
across tasks and facilitating information retrieval for post-processing
and reporting.
Clinical data - NLME models:
The main target of DDMoRe are clinical trials data and the natural choice
are the nonlinear mixed effect models (NLME) commonly used for the
analysis of longitudinal population data. The NLMEs are the mathematical
backbone of PharmML (REF1/2) determining its structure and scope.
The main estimation approach is the Maximum Likelihood Estimation
with the most popular algorithms being FO, FOCE, FOCE-I or EM type algorithms
(MCEM, MCPEM, SAEM) and others. See REF1/2 for overview.
Scope of PharmML:
1. Continuous data models, called structural model, can be defined as a system
of ordinary differential equation (ODE), delay differential equations (DDE)
and/or algebraic equations.
For encoding of compartmental pharmacokinetic models one can use
alternatively so called 'PK macros’, a system allowing for model formulation
without equations.
2. Discrete data models - covered are count, categorical, time-to-event data
models. Markov-type dependencies can be defined and examples had been
tested in connection with count and categorical models.
3. Parameter model allowing for implementation of virtually any parameter
type used in the NLME models, so called Gaussian model type or alternatively
formulated as equation.
4. Covariate model for integrating discrete or continuous covariates. The latter
can be transformed, interpolated or their distribution defined using UncertML
or our own ontology/knowledge base with more the 80 distributions and alternative
parameterisations.
5. Nested hierarchical variability model capable of expressing complex random
error structures.
6. Observation model with flexible residual error model supporting untransformed
or transformed continuous data.
7. Trial design model allowing for definition of many common design
patterns, drug administration types and encoding of experimental data
needed for typical simulation or estimation tasks, such as dosing, observations
and covariates.
8. Optimal Experimental Design support extends trial design with design
spaces on every model elements.
9. Hierarchical models/Bayesian inference is possible via assignment of
distributions to any model parameter.
10. Modelling steps: Specification of how a mathematical model and the
associated trial design can be used with typical modelling tasks such as
estimation, simulation, design optimisation/evaluation.
Scope of SO:
It consists of seven sections with space-holders for storing of tool settings,
raw results, task information, estimation, model diagnostic, simulation and
optimal design results.
See REF3/4 for detailed/general overview of the PharmML language.
A PharmML tutorial paper is in preparation. SO is going to be release publicly
soon.
Links:
ddmore.eu & pharmml.org
References:
REF1. Lavielle, M. Mixed Effects Models for the Population Approach Models,
Tasks, Methods & Tools (Chapman & Hall/CRC Biostatistics Series 2014).
REF2. Bonate, P. Pharmacokinetic-Pharmacodynamic Modeling and Simulation.
Springer Science & Business Media, 2011.
REF3. Swat et al. Pharmacometrics Markup Language (PharmML) Language
Specification for Version 0.6. Jan 2015. URL:www.pharmml.org/documentation2
REF4. Swat et al. Pharmacometrics Markup Language (PharmML): Opening New
Perspectives for Model Exchange in Drug Development. CPT Pharmacometrics
Syst Pharmacol. 2015 Jun;4(6):316-9.
Best Regards, Maciej
---------------------------
Maciej Swat
EMBL-EBI
mjswat at ebi.ac.uk
ddmore.eu & pharmml.org
twitter.com/PharmML
More information about the PopModWkGrpIMAG-news
mailing list