Home Software Survey Result
MIDAS Software Survey Results


Description

In the fall of 2009 the software working group asked for responses to a number of questions in order to get a snapshot of the current state of the software catalog. In all, nineteen of these surveys were completed and returned. The following are observations and suggestions based on the feedback contained within.

Software Catalog


Nine of the surveys returned were of software models, eight were of frameworks and two were visualization toolkits. Each package has been developed in house, most of which are publicly available or will be made so at some point.

The packages are as follows:

Health Care Facility Network Micro and Macro Models

The Micro and Macro Models are two software frameworks that create network models of patient flow among health-care facilities in a region:

  1. Equation-based deterministic model
  2. Agent-based model with each agent representing a patient.

These incorporate disease models that can then model the spread and control of infectious diseases throughout the network.

Contact - Bruce Lee, Levent Yilmaz, Kim Wong and Richard Christie, University of Pittsburgh

Disease Control System "DiCon"

DiCon is a framework for distributed, parallel disease simulation with policy optimization, e.g., for determining best strategies for resource distribution (temporal release of antivirals, vaccines; spatial optimization, etc.). DiCon is both the name for an entire framework, consisting of a core program and reusable code fragments of recurring implementation patterns that allow for rapid simulator development, and also the name for the core program itself that distributes user-provided simulator programs among a computer cluster. As of November 2009, only the main simulator dispatcher and optimizer program exists. As soon as the rest of the framework is in place, this program will become the new DiCon "core" in the bigger DiCon framework.

Contact - Lauren Ancel Meyers, Sebastian Goll and Nedialko Dimitrov, University of Texas

ABM++_D

ABM++_D is a framework for implementing and running distributed agent based models on Linux clusters. ÊIt supports either discrete event or time step update methods.

Contact - Doug Roberts, RTI

Parameter Estimation Kit (PEK)

The Parameter Estimation Kit set of MATLAB functions to estimate the optimal parameters to fit a given model to a set of data on an epidemic, possibly using limited, incomplete, or on-going data.

Contact - Sky Pelletier, University of Pennsylvania

EpiFast

EpiFast is a model that simulates the spread of an infectious disease across a social network.

Contact - Stephen Eubank and Keith Bisset, Virginia Tech

Didactic

Didactic is a web-based GUI and middleware system that allows people who are not HPC experts to setup run, and analyze complex simulations using a variety of models. Currently, EpiFast is available as a model.

Contact - Stephen Eubank and Keith Bisset, Virginia Tech

EpiSimdemics

EpiSimdemics is an interaction based individual model of the spread of contagion over a social network, with an extremely configurable set of interventions.

Contact - Stephen Eubank and Keith Bisset, Virginia Tech

Indemics
Indemics is an interactive simulation environment that supports dynamically querying and steering simulations that conform to a particular application programmer interface, that of EpiFast.  It is especially designed to handle the kinds of data used and created by agent-based models of epidemics, including social network information and network-related interventions.

Contact Ð Stephen Eubank and Madhav Marathe, Virginia Tech

FluTE

FluTE is an agent-based influenza epidemic simulation model system, that has been parallelized in MPI and is written in C++.

Contact - Dennis Chao, Fred Hutchinson Cancer Research Center

Steven Riley's spatial simulator

Steven Riley's spatial simulator is a library and two applications. One application is for smallpox and the other for influenza. Each application has two executables. The first builds a synthetic population with a network structure parameterized to commuting data. The second executable runs the simulation. Transmission occurs by one of three routes - household, network or spatial.

Contact - Steven Riley, University of Hong Kong

Spatial Modeling Kit(SMK)

The Spatial Modeling Kit is a set of MATLAB functions to model the spatial-temporal spread of an epidemic using a user-defined kernel and farm community.

Contact - Sky Pelletier, University of Pennsylvania

Global Scale Agent Model(GSAM)

The Global Scale Agent Model is a distributed agent based epidemic model written entirely in JAVA. ÊThe platform is unique for two reasons: it can support a very large population due to its efficient use of memory; it executes runs in a very short period of time (a 300 million agent run takes about 10 minutes). ÊThe GSAM is designed to be easily adaptable to different diseases and different assumptions about agent behavior.

Contact - Jon Parker, Brookings Institution

Repast Simphony

Repast Simphony is a widely used free and open source agent-based modeling platform.

Contact - Michael North, Argonne National Lab

Pitt/RTI Model

The RTI Model software is a tool for simulating infectious disease spread throughout a synthetic population.Ê It incorporates a number of mitigation strategies including vaccination, place closure, and antiviral treatments.

Contact - Shawn Brown, University of Pittsburgh, Phil Cooley, RTI

FRED

FRED is a framework for developing agent based models for infectious disease and behavior modeling.

Contact - Shawn Brown and John Grefenstette, University of Pittsburgh

GAIA

GAIA is a visualization tool designed to take a variety of geographically resolved data and represent it on a geospatial map.Ê It is a library that will eventually be incorporated into other packages.

Contact - Shawn Brown, University of Pittsburgh

VisuaLyzer

VisuaLyzer (Visual + Analyzer) is a computational tool for the visual analysis of epidemiological data.  It is intended to provide an intuitive understanding of relationships between variables within large datasets through rapid, dynamic, and intelligent data visualization.  The tool also includes a few fundamental epidemiology statistics which can be applied to data.

Contact Ð David Reshef, Massachusetts Institute of Technology

TBD

This is a systems dynamics disease propagation model which is still in development using the open source R platform.

Contact - Diane Lauderdale and Vanja Dukic, University of Chicago, Charles Macal, Argonne National Lab

TBD

This is an agent-based disease propagation model which is still in development using the Repast Simphony platform.

Contact - Diane Lauderdale, University of Chicago, Charles Macal and Michael North, Argonne National Lab

Licensing


Sixteen of these packages already are or have plans to become publicly available at some point.  GSAM and VisuaLyzer currently have no plans of becoming publicly available while Didactic and EpiSimdemics plan to create a business entity in order to eventually license the code.

Dependencies


Most libraries used to build the software are open source, however some software such as DiCon will require a GNU GPL or a BSD style of license.

User Community


These packages are generally used by only one or a few people though the Repast Simphony has several thousand users.  In some cases such as Didactic, the use of the software is as easy as using a web browser.   In cases such as FRED, EpiSimdemics, EpiFast and GSAM, the end user will need to run simple unix or Java based scripts and will need to be familiar with simple batch job submission.  In the case of GAIA, the user will need an understanding of Unix and Google Earth.  In the case of PEK, SMK and the Macro Model, the understanding of MATLAB syntax and the ability to run MATLAB functions is a necessity.  In the case of ABM++_D, the user will need to have experience in ABM and distributed message passing application development.  Writing computer code in C++ is necessary in order to use the Micro Model, DiCon, the RTI Model and FluTE.

Development Environment


C++ is the most popular development language among the surveys that we have received.  The Micro Model, DiCon, ABM++_D, EpiSimdemics, EpiFast, FluTE, Steven Riley's spatial simulator and the RTI Model all have been developed using C++ to some extent. MPI has been used to develop packages such as DiCon, EpiFast and EpiSimdemics.  MATLAB was used to develop the Macro Model, PEK and SMK.

Computing System and Needs


Most of the software will run on Linux, Microsoft Windows or Apple OS platforms and generally don't require a large amount of processing power.Ê However, some of the models such as ABM++_D can be scaled up to run large simulations across thousands of processors.  In terms of storage, most projects require only a small amount of disk space for data, however the Health Care Facility Network Micro and Macro Models uses 100s of GBytes of space.

 

Visualization


Visualization techniques vary slightly across the packages.  Didactic, EpiFast and EpiSimdemics use custom built tools with the analysis done using R.  Steven Riley's spatial simulator is also using R.  FluTE is using a combination of ArcGIS and R.  The Micro and Macro Models, PEK and SMK are all using MATLAB.  GSAM produces it's own movie output as well as epidemic curves.  FRED and the RTI Model are using GNUPlot, Excel and GAIA which in turn uses Google Earth.  The VisuaLyzer is still in heavy development.

Documentation

Significant end user documentation only exists in a few instances such as the RTI Model, Repast Simphony, Didactic and EpiFast.

Future Plans

The future plans vary widely as one might expect.  Didactic, EpiFast, EpiSimdemics and Indemics are part of a suite of interrelated packages.  The plan for this suite is to add support for a more sophisticated run-time simulation-steering capability.  There are plans to merge PEK and SMK into one set of tools with a graphical front-end to create a straightforward system for working with discrete spatial-temporal models of disease epidemics.  There are plans to implement their "Coupled Contagion" work into the GSAM model.  The Repast Simphony is being regularly updated.  The features of the RTI Model are to be incorporated in the FRED framework, which will then be coupled with a visualization package.  Increasing network detail, augmenting transmission model, and introducing intervention modules are in the works for the Micro and Macro Models.  Steven Riley plans to continue to work on the spatial simulator for his own research.  Upgrades will include improvements to the comments and the usability and to broaden it's availability.  Also he plans to slowly implement various levels of parallelization using MPI.  Finally, the DiCon developers would like to extend the framework with reusable code fragments of recurring implementation patterns that allow for rapid simulator development.  The current simulator dispatcher and optimizer program then becomes the "core" in the new DiCon framework.

Observations

The groups do not seem to be constrained by computing hardware at all, but rather are comfortable running on the platforms that they are currently using.  Many of the projects mention that improvements would be made simply as their time would allow. Also, the individual groups seem to have a good grasp on the eventual direction of the codes in terms of its availability to the public as well as what their intended user group will look like.  Documentation is an area that will require the attention of many of the groups in the future.

 

 



Last Updated on Wednesday, 05 May 2010 11:03
 

User Login