Download as pdf or txt
Download as pdf or txt
You are on page 1of 30

Computational forecasting of 

infectious disease dynamics

Daniel Alejandro Gonzalez Bandala
´
danielgoba84@gmail.com
Contents
Introduction
Bioinformatics
• Datasets
Artificial Intelligence (AI)
• Data Mining (DM)
• Artificial Neural Networks (ANN)
Hypothesis
Objectives
Courses
Timetable
Introduction
Zoonotic diseases have increased their 
impact in human health. 
Bats are hosts of many zoonotic 
diseases. 
These viruses apparently cause little to 
none pathologies to bats, allowing them 
to live infected. 
(Calisher, C. H., Childs, J. E., Field, H. E., Holmes, K. V., & Schountz, T.(2006). Bats: important reservoir hosts of emerging viruses.
Clinical microbiology reviews, 19 (3), 531–545.)
(Plowright, R. K., Eby, P., Hudson, P. J., Smith, I. L., Westcott, D., Bryden, W. L., . . . Martin, G., et al. (2015). Ecological dynamics of
emerging bat virus spillover. In Proc. r. soc. b (Vol. 282, 1798, p. 20142124). The
Royal Society.)
(Wang, L. & Crameri, G. [G]. (2014). Emerging zoonotic viral diseases. Rev sci tech Off int Epiz, 33 (569-81). )
Table 1. Zoonotic viruses causing disease in human and their bat reservoir hosts. 
(Ranjan, K., Prasad, M., & Prasad, G. (2016). BATS: CARRIERS OF ZOONOTIC VIRAL AND EMERGING INFECTIOUS 
DISEASES. Journal of Experimental Biology, 4, 3S.)
Figure 1. Example of possible vectors of infection from bats to humans.
(Wang, L. F., & Crameri, G. (2014). Emerging zoonotic viral diseases. Rev sci tech Off int Epiz, 33(569-81))
Figure 2. Enabling conditions for Hendra virus spillover. 
(Plowright, R. K., Eby, P., Hudson, P. J., Smith, I. L., Westcott, D., Bryden, W. L., ... & Tabor, G. M. (2015, January). Ecological 
dynamics of emerging bat virus spillover. In Proc. R. Soc. B (Vol. 282, No. 1798, p. 20142124). The Royal Society.)
Bats are present in all the continents,
except the Antarctic.
They are second on global distribution
of mammals only after humans
México holds 15% of the total diversity
of bats.
This research proposal is part of the project
“Establecimiento de la Red de Vigilancia de
Enfermedades Virales Emergentes (rVEVE)”*.
Proposes the development of molecular and
bioinformatic tools that would help to detect,
monitor, alert and even predict or simulate the
emergence of an infectious disease outbreak.
*Proyectos de desarrollo científico para atender problemas Nacionales, convocatoria 
2016, CONACYT
This proposal focuses on finding patterns
in the bats’ genetic information data set
and Google trend topics that point towards
new possible emergent zoonotic diseases in
order to fight them before they spill to
humans.
Workgroup
Laboratorio de Genómica Viral y Humana 
(LGVH) at Medicine Faculty of the UASLP.
Engineering Faculty (FI) of the UASLP.
Comité Estatal de Fomento y Protección Pecuaria 
(CEFPP).
Laboratorio de Inmunología Molecular (LIM) at 
the Biomedic Research Center of the UAdeC in 
Torreón.
The Instituto de Ciencia SA de CV (IC) from 
Torreón, Coahuila.
M.C.C. Daniel Alejandro González Bandala
Ph.D student candidate.

Dr. Juan Carlos Cuevas Tello
Main adviser at Engineering Faculty, UASLP.

Dr. Christian A. García Sepúlveda
Secondary adviser at Medicine Faculty, UASLP.
Bioinformatics
The use of technologies in the
organization, management and analysis of
biologic data.
To find possible zoonosis outbreaks, it is
necessary to analyze several bioinformatic
data from bats along with reports of
suspicious cases of emergent infectious
diseases.
Data sets
MIDASmap (Mexican Infectious Disease Analysis and Surveillance map)
An online interactive map which will help to follow up the
georeferenced information gathered.
Plot virological screening information from bats.
Plot alert information from not reported emergent viral
disease outbreaks or suspicious cases.
Plot results from Google trends data mining to identify pre-
hospital outbreaks.
Allow simulation and prediction of emergent infectious
diseases from forecasting algorithms.
MIDASmap
Additionally, a cellphone app will be created to
allow fast remote report creation of suspicious
cases of emergent infectious diseases to feed
MIDASmap.
Figure 3. MIDASmap prototype, temporary URL: healthmap.tk
Data sets
Google trending topics
Monitoring keywords or search terms used to trigger an
epidemiological alert, once some defined threshold is exceeded
pointing to users searching about clinical manifestations of
emergent infectious diseases.

These results will be correlated with information gathered in


MIDASmap, allowing the prediction of emergent infectious
diseases.
Artificial Intelligence (AI)
A set of computational tools will be
developed for the gathering, mining and
forecasting of data.
This research will focus on data mining 
an artificial neural networks forecasting 
models
Data mining (DM)
Data mining is a step in the Knowledge
Discovery in Databases (KDD) process
that consists of applying data analysis and
discovery algorithms that produce a
particular enumeration of patterns (or
models) over the data.

(Fayyad, U., Piatetsky-Shapiro, G., & Smyth, P. (1996). From data mining to knowledge discovery in databases. AI 
magazine, 17(3), 37.)
Figure 4. Steps of the KDD process. 
(Fayyad, U., Piatetsky-Shapiro, G., & Smyth, P. (1996). From data mining to knowledge discovery in databases. AI 
magazine, 17(3), 37.)
Data mining involves an integration of
techniques from multiple disciplines such as:
Data base and data warehouse technologies
Statistics
Machine learning
Pattern recognition
Neural networks
Image and signal processing
Spatial and temporal data analysis
Etc.
Artificial Neural Networks (ANN)
ANN have long been used for forecasting, originally, with
much uncertainty.
In principle, they learn from example, and capture subtle
functional relationships among data.
Even if the aforementioned relationships are unknown, ANN
are capable of performing nonlinear modeling without prior
knowledge about the relationship between input and output
data.
ANN can identify and learn correlated patterns between input
data sets and corresponding target values. This technique is
ideally suited for modeling imprecise and noisy data which may
desirable feature for kind of data being modeled.
(G. Zhang, B. Patuwo, and M. Hue, "Forecasting with artificial neural networks: The state of the art, " International 
Journal of Forecasting, vol. 14, 1998.)
(G. K. Jha, P. Thulasiraman and R. K. Thulasiram, "PSO based neural network for time series forecasting," 2009 
International Joint Conference on Neural Networks, Atlanta, GA, 2009, pp. 1422-1427. doi:  10.1109/IJCNN.2009. 
5178707 ) 
ANN are more general and flexible modeling technique for
forecasting.

The standard back propagation training algorithm for neural


networks exhibits slow convergence, local minima, and lack of
robustness.

There does not exist any algorithm which can guarantee the global
optimal solution for a general non-linear optimization problem in a
stipulated time.

(D. Rumelhart, J. McClelland, and the PDP Group, Parallel Distributed Processing, Explorations in the 
Microstructure of Cognition. Cambridge: MIT Press, 1986, Vol. I.)
(G. Zhang, "Avoiding pitfalls in neural network research, " IEEE Transaction on Systems, Man, and Cybernetics-Part 
C: Applications & Reviews, vol. 37, 2007.)
Artificial Neural Networks (ANN)
ANN have high capabilities in a great
number of everyday problems.
They are widely used for:
Pattern recognition
Function approximation
Prediction/forecast
Optimization among other applications.
(Haykin, S. (1994). Neural networks, a comprehensive foundation.)
(Bishop, C. M. (1995). Neural networks for pattern recognition. Oxford university press.)
Spiking Neural Networks (ANN)
SNN work with train of pulses as inputs, this
makes them resemble the behavior of their
biological counterparts.
Considered the third generation of ANN.
Their performance has proved better than
sigmoidal neural networks.
Supervised training with risk of stucking on
local minimums.
(Maass, W. (1997). Networks of spiking neurons: the third generation of neural network models. Neural networks, 
10(9), 1659-1671.)
(Altamirano, J. S. (2015). Comparación de Algoritmos Metaheurísticos Aplicados al Entrenamiento de Redes 
Neuronales Pulsantes. (Tesis de Maestría). Instituto Tecnológico Nacional de México.)
Deep Neural Networks (ANN)
DNN can be defined as the ANN with multiple
hidden layers.
These additional layers add greater recognition
capabilities than the common ANN.
The multiple layers build up growing levels of
abstraction.
Once trained they have a great performance and are
resistant to input noise
(Cuevas-Tello, J. C., Valenzuela-Rendón, M., & Nolazco-Flores, J. A. (2016). A Tutorial on Deep Neural 
Networks for Intelligent Systems. arXiv preprint arXiv:1603.07249.)
(Deng, L., Hinton, G., & Kingsbury, B. (2013, May). New types of deep neural network learning for speech 
recognition and related applications: An overview. In 2013 IEEE International Conference on Acoustics, Speech and 
Signal Processing (pp. 8599-8603). IEEE.)
Hypothesis
Artificial intelligence algorithms allow the
computational forecasting of infectious disease
dynamics.
Objectives
Main objective

To find patterns in the bats’ genetic information


data set and Google trend topics that point
towards new possible emergent zoonotic diseases
in order to fight them before they spill to
humans.
Objectives
Specific objectives
To study the information contained in MIDASmap.
To monitor Google trend topics and the correlation
with the MIDASmap information.
To study and develop data mining algorithms.
To research computational forecasting models.
To measure the significance of results statistically.
Courses
Machine Learning
Engineering Faculty UASLP.

Bioinformatics
Engineering Faculty UASLP.

Research Methodology
Engineering Faculty UASLP.
Timetable
Semester 1 (2017) Semester 2 Semester 3 Semester 4 Semester 5 Semester 6 Semester 7 Semester 8
(2018) (2019) (2020)
Thesis progress Seminar1 Seminar 2 Seminar 3 Seminar 4 Seminar 5 Seminar 6 Writing Up Writing UP
Courses Machine Learning Neural Networks Bioinformatic
*Studying MIDASmap and Google Trends
*Literature Review of computational       
forecasting models

*Studying MIDASmap and Google Trends
*Literature Review of computational 
forecasting models
*Elaborate thesis proposal
*Developing of data mining algorithms
*Developing of artificial neural networks 
algorithms

*Integrating MIDASmap and Google Trends 
with AI algorithms
*Study correlation between MIDASmap and 
Google Trends
Middle Term exam

*Measure significance of AI algorithms
*Start writing up of journal paper
*Research stay in UK
*Submit Journal paper
*Analyzing final results and their statistical 
study
*Previous exam
*Grade exam

You might also like