how can users benefit from the ddmore common language
play

How can users benefit from the DDMoRe common language standard and - PowerPoint PPT Presentation

How can users benefit from the DDMoRe common language standard and interoperability framework? World Conference on Pharmacometrics 2016 Phylinda LS Chan, Lutz O Harnisch, Peter A Milligan, Mike K Smith On behalf of the DDMoRe consortium Typical


  1. How can users benefit from the DDMoRe common language standard and interoperability framework? World Conference on Pharmacometrics 2016 Phylinda LS Chan, Lutz O Harnisch, Peter A Milligan, Mike K Smith On behalf of the DDMoRe consortium

  2. Typical modelling problems today Limited access to existing modelling knowledge Non-compatible software languages Waste of time and resources, more errors working in a non-integrated/non-compatible environment 2

  3. DDMoRe Consortium in a nutshell 5.5 years Start 23 million 26 March Until Aug € budget partners 2011 2016 3

  4. 4

  5. DDMoRe’s Vision Improve quality, efficiency and cost effectiveness of Model-Informed Drug Discovery and Development (MID3) and therapeutic use 5

  6. DDMoRe Platform: “Step by step” Scientific Question  Task & Workflow “User perspective “ (June 2016)

  7. DDMoRe Platform: “Step by step” Executable files “Task manager” 8 “User perspective “ “Integration manager” “Translator” (June 2016)

  8. DDMoRe Platform “User perspective “ “User perspective “ (July 2016) (July 2016)

  9. PharmML & SO – Big Picture Pharmacometric Markup Language & Standardized Output

  10. MDL – ProbOnto 11

  11. Model Description Language Structure 12

  12. Model Description Language Structure 13

  13. Modularity – example workflow ▪ Estimation = Data + Parameters + MODEL + Monolix Task Properties ▪ Bayesian estimation = Data + Priors + MODEL + BUGS Task Properties ▪ VPC = Data + Final Parameters + MODEL + NONMEM Task Properties ▪ Prediction / simulation = Design + Final Parameters + MODEL + Simulation Task Properties ▪ Optimal design / evaluation = Design + Final Parameters + MODEL + PFIM / PopED Task Properties 14

  14. MDL – How is it different?  Definition of dataset variables and their use in the model DATA_INPUT_VARIABLES{ ID : { use is id } TIME : { use is idv } WT : { use is covariate } AGE : { use is covariate } SEX : { use is catCov withCategories{female when 0, male when 1} } AMT : { use is amt, variable = GUT } DVID : { use is dvid } DV : { use is dv, define={1 in DVID as CP_obs, 2 in DVID as PCA_obs} } MDV : { use is mdv } }# end DATA_INPUT_VARIABLES 15

  15. MDL – How is it different? ▪ To support interoperability , some relationships need to be more explicit (when they can be)... 𝛾 𝐷𝑀 𝛾 𝐷𝑀 log 𝐷𝑀 𝑗 = 𝑚𝑝𝑕 𝐷𝑀 × 𝑋𝑈 𝐷𝑀 𝑗 = 𝐷𝑀 × 𝑋𝑈 × 𝑓 𝜃 𝑗 + 𝜃 𝑗 70 70 INDIVIDUAL_VARIABLES { CL : { type is linear , trans is ln, pop = POP_CL, fixEff = [{ coeff =BETA_CL_WT, cov =logtWT}] , ranEff = ETA_CL ) ... } # end INDIVIDUAL_VARIABLES 16

  16. MDL – Some things are easier… ▪ MDL uses the ProbOnto knowledge base for statistical distributions used in Random Variable Definitions. MODEL_PREDICTION{ lnLAMBDA = ln(BASECOUNT) + BETA*CP LAMBDA = exp(lnLAMBDA) } RANDOM_VARIABLE_DEFINITION(level=DV){ Y ~ Poisson1( rate=LAMBDA) } OBSERVATION{ :: { type is count, variable = Y} }# end ESTIMATION 17

  17. MDL-IDE Manages all modelling tasks Project explorer Model/script editor where standardized with advanced editing outputs are and syntax checking accessible features Task console (R) 18

  18. Use Cases Demonstrate functionality, model features • Warfarin Pop PK used as the basis. • UseCase10 – 2 distribution compartments • UseCase1 – Structural model with differential • UseCase11 – Poisson count equations • UseCase2 – Analytical solution • UseCase12 – Categorical outcome • UseCase3 – Joint model of PK and PD (>1 outcome) • UseCase13 – Binary outcome • UseCase4 – IV and oral administration • UseCase14 – Time to event (Exact time of event known) • UseCase5 – Covariate models including categorical • UseCase15 – log-transformed DV covariate • UseCase16 – BLQ handling • UseCase6 – Correlation between V, CL, KA • UseCase17 – Steady State PK • UseCase7 – Structural model specified by • UseCase16 – Interpolation of covariates compartments • UseCase19 – L2 handling • UseCase8 – Between-occasion variability • UseCase20 – Transit compartment for absorption • UseCase9 – Infusion rates • UseCase21 – Mixture models • UseCase22 – Complex PK (>1 absorption compartment) • UseCase23 – Conditional observation model Interoperable UseCase Valid MDL In progess 19

  19. Typical modelling problems addressed by DDMoRe products today Limited access to existing modelling knowledge Non-compatible software languages Waste of time and resources, more errors working in a non-integrated/non-compatible environment 21

  20. In Summary  In support of MID3, DDMoRe products provide a vital improvement to transparency in model informed decision making, enhancing knowledge-sharing and scientific communication.  Deliverables of the DDMoRe project provide • a quantitative framework for prediction and extrapolation, • centred on knowledge and inference generated from integrated models of compound, mechanism and disease level data, • hence improving the quality, efficiency and cost effectiveness of decision making. 22

  21. More information about DDMoRe  Official webpage: http://ddmore.eu/ • For updates on project, products, news, newsletters, publications, blog, events, forums, FAQ, etc. • Follow DDMoRe: Twitter  Videos: • DDMoRe vision: https://www.youtube.com/watch?v=zxsNOewJ84g • Podcasts: https://www.youtube.com/watch?v=kgGw4uc2hmw • MDL IDE installation: https://www.youtube.com/watch?v=kgGw4uc2hmw • DDMoRe: an introduction and demo (ISOP workshop): https://www.youtube.com/watch?v=7FmrPKAhFKM 23

  22. How do I get started?  Drug and Disease Model Repository http://repository.ddmore.eu/  Interoperability Framework • Installation and User Cases: http://ddmore.eu/beta-release-interoperability-framework • MDL User Guide: http://www.ddmore.eu/instructions/mdl-user-guide  DDMoRe Foundation • Contact the Board members: Marylore Chenel, Lutz Harnisch, Mats Karlsson, Paolo Magni, Peter Milligan. 24

  23. Participants are a unique combination of model builders, model users, software developers and teachers 25

  24. THANK YOU… 26 On behalf of the DDMoRe consortium

Recommend


More recommend