11 Changelog
11.1 v1.0 - 19 August 2016
Public release of MDL, PharmML, DDMoRe Interoperability Framework Standalone Execution Environment (SEE)
MDL v1.0 MDL User Guide for public release
The MDL Documented in the MDL User Guide hosted on Github is the latest release (currently v1.0) but the documentation has been updated where necessary. The version linked above is a snapshot of documentation for the IMI DDMoRe project deliverable at the close of the project.
- PharmML v0.8.1
11.1.1 Scope
The primary focus of MDL in this release is translation to valid PharmML, rather than conversion to target software. The previous release was primarily concerned with demonstrating interoperability across key software targets. In this version of MDL there may be MDL features which are not supported in conversion from PharmML to certain target software, but which are valid for model description and which generate valid PharmML. The aim is to widen the scope of models which can be encoded in MDL and generate PharmML for uploading to the DDMoRe repository and for future interoperability. Translation of these models to target software will follow with updates to the interoperability framework converters.
The changes to MDL since draft 7 (v0.7) enable integration of the Prior and Design Objects and improved validation of MDL giving increased confidence in generation of valid PharmML. In order to facilitate this, certain changes to syntax have been made that are NOT backwards compatible. This is regrettable since it means that existing models required changes. We do not make these changes lightly.
11.1.2 CHANGES v0.7 to v1.0
The follow are critical changes which break backward compatibility with v0.7:
11.1.2.1 Data Object
DECLARED_VARIABLES
block must now have type assigned to variables to enable validation of variable types between MDL objects, particularly Design Object variables.
11.1.2.2 Parameter Object
Correlations and covariances between parameters are now specified in the Model Object and must be named parameters. This is to facilitate specification of priors on these parameters.
Users should not specify the type of variability definition (
type is sd
,type is var
,type is corr
,type is cov
) forVARIABILITY
parameters.The variability, covariance or correlation type is specified and used in theRANDOM_VARIABLE_DEFINITION
block in the Model Object where these parameters are defined.
11.1.2.3 Model Object
Left hand side transformations for [
INDIVIDUAL_PARAMETERS
] are no longer valid. These were felt to be confusing.Right hand side functions for [
INDIVIDUAL_PARAMETERS
] andOBSERVATION
definitions are now list definitions with the matching type to the function. This, combined with conditional statements allows more flexibility in parameter andOBSERVATION
definitions.Non-continuous outcomes (binary, count, categorical) must be defined as RANDOM_VARIABLE_DEFINITION(level=DV){ … } and then the variable defined assigned as an anonymous list in the
OBSERVATION
block. This is to ensure that outcome variables are always defined with variability at the DV level. (This is implicit in continuous outcomes due to definition of residual error at the DV level).Arbitrary equations defining the outcome variable are not allowed in the
OBSERVATION
block. Use “type is userDefined” instead.
11.1.3 NEW FEATURES in v1.0
11.1.3.1 General
Support for vectors and matrices
Support for conditional assignment to lists.
11.1.3.2 Data Object
Support for model input variables passed from the Data Object with “use is variable”. This equates to “regressor” type inputs to models.
Support for
DATA_DERIVED_VARIABLES
where dose amounts and dose times can be derived from data columns which are being used otherwise asuse is amt
oruse is idv
.
11.1.3.3 Parameter Object
No new features. See changes section above.
11.1.3.4 Model Object
Support for definition of parameters in
RANDOM_VARIABLE_DEFINITION
i.e. definition of CL ~ Normal(mean=POP_CL, sd=PPV_CL) for subsequent use in [INDIVIDUAL_PARAMETERS
]. This combined with support for ProbOnto definitions allows the user to define multivariate distributions of parameters, mixture distributions etc.Support for ProbOnto distributions in
RANDOM_VARIABLE_DEFINITION
.Support for combination of compartment definitions with differential equations.
Support for userDefined specification of the relationship between model predictions and residual error random variables in the
OBSERVATION
block.
11.1.3.5 Task Properties object
- Support for target software specific Task Properties object.
11.1.3.6 MOG Object
- Support for specifying problem statement and model name in the
INFO
block.
11.1.3.7 Prior Object
The Prior Object is a new block for v1.0
11.1.3.8 Design Object
The Design Object is a new block for v1.0
11.2 v0.7 - 11 December 2015
First public release of MDL
MDL v0.7 MDL User Guide
PharmML v0.6