The Reference Model
Last updated: May 2017
Participated in following Mt Hood Diabetes Challenge meetings: 2012, 2014, 2016.
Publicly accessible?: Aspects of the model are subject to a patent.The Reference Model code is currently unreleased to the public. However, the MIcro Simulation Tool (MIST) that drives the model is free software with some versions accessible through the following links:
Is the model continuing to be developed? Yes
The Reference Model for Disease Progression can be described in several ways:
An ensemble model that acts like an assembly plant that assembles one model from multiple other models
A league of disease models where models can compete and cooperate to better explain observed phenomena
A validation model that shows how well a certain model combination describes certain populations
A knowledge accumulator that gathers both observed population data and modeling assumptions
The Reference Model employs High Performance Computing (HPC) to combine computational building blocks to best fit multiple populations. Those computational building blocks can be either other published models or assumptions. The Reference Model therefore automates and extends the process modelers followed in previous Mount Hood validation challenges – it centralizes validation of multiple public models against the populations used in Mount Hood challenges since 2010 and many others.
The Reference Model now employs an assumption engine that allows computational components to compete and cooperate to find better fitting model combination. The Reference Model is composed from multiple competing models, therefore its results show our mutual understanding of disease progression.
The Reference Model is a good way to cross reference information to find out pieces of information and assumptions that fit together, and allow competition against accumulated known data to guide our perception. HPC is a key to those capabilities and it provided using capabilities of the MIcro Simulation Tool (MIST).
MIST also provides advanced population generation techniques using Evolutionary Computation. This allows the model to use publicly available data such as clinical trial reports, including the demographic statistics and the outcomes measures. The Reference Model is therefore capable of gathering more information since it can access data that otherwise would be restricted from sharing. This distinguishes the model from other models that are typically constructed using one trial as a major data source – The Reference Model is built from merging information from multiple sources and can be considered as a knowledge repository.
The Reference Model has recently been enhanced with an interface that allows it to read information from ClinicalTrials.Gov while maintaining tractability and reproducibility.
Funding source for model development
The Reference Model and MIST are self funded without financial support.
However, early funding allowed development of tools that enabled the model. The MIcro Simulation Tool (MIST) that is based on IEST. The IEST GPL disease modeling framework was initially supported by the Biostatistics and Economic Modeling Core of the MDRTC (P60DK020572) and by the Methods and Measurement Core of the MCDTR (P30DK092926), both funded by the National Institute of Diabetes and Digestive and Kidney Diseases. The modeling framework was initially defined as GPL and was funded by Chronic Disease Modeling for Clinical Research Innovations grant (R21DK075077) from the same institute.
Barhak J. The Reference Model for Disease Progression. SciPy 2012, Austin Tx, 18-19 July 2012.
Barhak J, The Reference Model for Disease Progression – Data Quality Control. SummerSim 6-10 July 2014, Monterey CA.
Barhak J., The Reference Model for Disease Progression uses MIST to find data fitness. PyData Silicon Valley 2014 held at Facebook Headquarters:
Barhak, J. Modeling Clinical Data from Publications, SpringSim 2015. April 12 - 15, Alexandria, VA, USA. Presentation
Barhak, J. The Reference Model Uses Modular Population Generation! Object Oriented Population Generation on the Fly with MIST. IMAG Multiscale Modeling (MSM) Consortium Meeting 9-10 September 2015.
Barhak J. The Reference Model uses Object Oriented Population Generation. SummerSim 2015 July 26-29, Chicago IL, USA.
Barhak, J. The Reference Model for Disease Progression and Latest Developments in the MIST, PyTexas 2015. College Station, TX, 26-Sep-2015. Presentation
Barhak, J. Garrett, A. Pruett, W. A. Optimizing Model Combinations, MODSIM world 2016. 26-28 Apr, Virginia Beach Convention Center, Virginia Beach, VA. Presentation
Barhak J. The Reference Model for Disease Progression and Latest Developments in the MIST, PyTexas 2015. College Station, TX, 26-Sep-2015.
Barhak J. The Reference Model Predicting Life Expectancy for Diabetics, MODSIM world 2016. 26-28 Apr, Virginia Beach Convention Center, Virginia Beach, VA.
Barhak J. The Reference Model for Disease Progression Combines Disease Models. I/IITSEC 2016, 28 Nov – 2 Dec Orlando Florida. USA.