Introduction - DigitalCommons@UNO

Joseph Devlin
RCAF 2017
Ruiz-Moreno CHIKV Model
Mentor: Dr. Kathryn Cooper PHD.
University of Nebraska Omaha Bioinformatics
March 3, 2017
Agenda
• Introduction
• CHIKV
• SEIR Methodology
• Goal
• Methods
• Creation of Control Data
• Code to Compare Model Data to PAHO data
• Initial/Continual Testing of the Model
• Conclusion
• Continuation of Model Testing
Introduction
• Project is an extension of the paper, “Modeling Dynamic Introduction
of Chikungunya Virus in the United States”.
• Creation of a CHIKV SEIR model – common use in epidemiology for
disease prediction
Introduction
• CHIKV
• Virtually no cure, no vaccine, and no specific treatment
• Best treatment seems to be symptom control
• Vector borne disease
• Carried by tropical mosquitos
• Optimal for SEIR methodology
Introduction
• SEIR Methodology
• The model is a set of variables and parameters that all have impact on each
other for a final outcome given a certain set of starting parameters
• Follows the population under certain states:
• S = Susceptible
• E = Exposed
• I = Infected
• R = Recovered
Goal
• To work out a way to modify the model to increase accuracy
• Our focus was on the infected populations
• Accomplished by:
• Creation of files to compare model data to – control data
• Code in order to compare model data to known data
• A set of steps in order to continually modify and test the model
Methods - Control Files
• Creation of files to compare model data to
• Source: PAHO CHIKV statistic data
• Uncertainty: Not all countries reported each epidemiological week – not all
counts can be guaranteed accurate
• To account for this, if data was not reported, the data for that week was taken from the
previous reported week
Methods – Control Files
• Full completion resulted in epidemiological data for 26 countries that
reported to PAHO for control comparison
Methods – Comparison
• Comparison of model data vs. control data uses equation below to
measure accuracy of model data
• 𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦 𝑆𝑐𝑜𝑟𝑒 =
1
𝑁𝑐 𝑁𝑊
𝑐,𝑤 −
∗ −𝑥
2
(𝑥𝑐,𝑤
𝑐,𝑤 )
2𝜎𝑐,𝑤 2
• This was put into practice using “Perl” code
• Process:
• Takes two input files (PAHO control and chosen model data)
• Runs week by week comparison (PAHO week 1 vs. Model week 1 and so on)
• Outputs a single score for accuracy comparison
Methods – Model Data/Testing
• Model parameters are programmed using “R”
• Model input: Country Name, Population, S, E, I, R, and G (mosquito
population)
• Model Output: Graph of S, E, I, and R variables and another .CSV file
which we used to compare to control data
• Chosen and created model data for comparison:
• Individual month data using their own known initial variables
• Continuous month-to-month run (variables come from models predicted
outcomes from previous month)
• i.e. Octobers initial variables came from September’s output
Methods – Model Data/Testing
• Put it all together to get a readable outcome - Accuracy score
• Usability of scores: tracks progress if model modification made was
beneficial or detrimental
Conclusion
• Set of steps and materials to continually modify and improve the
model