Event Abstract

Back to Event

Spatial Machine Learning Approach for Risk-Based Surveillance of Bluetongue Virus Serotypes in Europe

Moh A. Alkhamis1, 2*, Nicholas M. Fountain-Jones3, Andres M. Perez3 and Jose M. Sanchez-Vizcaino4
  • 1 Kuwait University, Kuwait
  • 2 Faculty of Public Health, Kuwait University, Kuwait
  • 3 Department of Veterinary Population Medicine, College of Veterinary Medicine, University of Minnesota, United States
  • 4 VISAVET Health Surveillance Centre (UCM), Spain

Bluetongue virus (BTV) is a notifiable midge-borne pathogen of ruminants that has led to severe economic losses across Europe (Maclachlan et al. 2015, Rushton and Lyons 2015). BTV is classified into 27 antigenically distinct serotypes (Jenckel et al. 2015) with at least 8 serotypes responsible for over 11 000 outbreaks in Europe since the year 2000. Most notable outbreaks were reported in France, Italy, Spain, Swaziland, and the Balkans, which were caused mainly by serotypes 1, 4 and 8 (Breard et al. 2007, Corbiere et al. 2012, Durand et al. 2010, Nicolas et al. 2018). While outbreaks are continuously reported across Europe on an annual basis, the need for improved risk-based surveillance activities is essential for reducing the impact of BTV on the region. The objective of this study was to model the spatial distributions of BTV outbreaks with particular emphasis on serotypes 1,4, and 8, and to identify their unique environmental requirements in Europe using a machine learning statistical framework. Our data compromised 9296 unique locations for outbreaks reported between 2000 and 2018 in Europe. Environmental high-dimensional data included a total of 23 variables and mainly compromised climate, animal densities, Culicoides spp. abundance and landcover. We applied three relatively popular and robust machine learning models, including random forest (RF), support vector machine (SVM), and boosted regression trees (BRTs), and compared their predictive importance. We also compared the predicted spatial distribution and its most important environmental variables of all BTV outbreaks versus serotypes 1, 4, and 8 alone. Based on a 10-fold-cross-validation approach, all models performed very well with area under the curve (AUC) scores higher than 0.90. However, the consistently highest predictive model was achieved using RF with an AUC score of 0.99. Using all BTV presence data, the highest predicted spatial risk (probability > 0.8) was constrained within France, Italy, Swaziland and the Balkans where the majority of the outbreaks were reported (Fig 1B). Similarly, the highest predicted spatial risk was constrained in Italy for serotype 1, in northern Italy and the Balkans for serotype 4, and in France and Swaziland for serotype 8 (Figs 1C, D and E). Based on the mean decrease Gini estimated by the RF model, goat density and mean diurnal range were the most important predictors for the spatial risk of all BTV serotypes as well as serotype 4 alone (Maclachlan et al. 2009). However, buffalo density and annual mean temperature were the most important predictors for serotype 1 (Lorusso et al. 2014), while cattle density and abundance of Culicoides spp. were the most important environmental predictors for serotype 8 (Zanella et al. 2012). We provided a unique approach for the use of machine learning models on big data for risk-based surveillance in Europe (i.e., serotype focused). We mapped the predicted high-risk areas of BTV outbreaks with high accuracy. Furthermore, we identified the most important host species for each serotype in combination with other risk factors, including climate and vector abundance. The results will help in guiding current BTV surveillance activities in Europe, which subsequently will improve control efforts and reduce the devastating impact of new outbreaks on the regional animal resources.

References

Breard E, Sailleau C, Nomikou K, Hamblin C, Mertens PP, Mellor PS, El Harrak M, Zientara S. 2007. Molecular epidemiology of bluetongue virus serotype 4 isolated in the Mediterranean Basin between 1979 and 2004. Virus Res. May;125:191-197. Epub 2007/02/07. Corbiere F, Nussbaum S, Alzieu JP, Lemaire M, Meyer G, Foucras G, Schelcher F. 2012. Bluetongue virus serotype 1 in wild ruminants, France, 2008-10. J Wildl Dis. Oct;48:1047-1051. Epub 2012/10/13. Durand B, Zanella G, Biteau-Coroller F, Locatelli C, Baurier F, Simon C, Le Drean E, Delaval J, Prengere E, Beaute V, et al. 2010. Anatomy of bluetongue virus serotype 8 epizootic wave, France, 2007-2008. Emerg Infect Dis. Dec;16:1861-1868. Epub 2010/12/03. Jenckel M, Breard E, Schulz C, Sailleau C, Viarouge C, Hoffmann B, Hoper D, Beer M, Zientara S. 2015. Complete coding genome sequence of putative novel bluetongue virus serotype 27. Genome Announc. Mar 12;3. Epub 2015/03/15. Lorusso A, Sghaier S, Ancora M, Marcacci M, Di Gennaro A, Portanti O, Mangone I, Teodori L, Leone A, Camma C, et al. 2014. Molecular epidemiology of bluetongue virus serotype 1 circulating in Italy and its connection with northern Africa. Infect Genet Evol. Dec;28:144-149. Epub 2014/09/23. Maclachlan NJ, Drew CP, Darpel KE, Worwa G. 2009. The pathology and pathogenesis of bluetongue. J Comp Pathol. Jul;141:1-16. Epub 2009/05/30. Maclachlan NJ, Mayo CE, Daniels PW, Savini G, Zientara S, Gibbs EP. 2015. Bluetongue. Rev Sci Tech. Aug;34:329-340. Epub 2015/11/26. Nicolas G, Tisseuil C, Conte A, Allepuz A, Pioz M, Lancelot R, Gilbert M. 2018. Environmental heterogeneity and variations in the velocity of bluetongue virus spread in six European epidemics. Prev Vet Med. Jan 1;149:1-9. Epub 2018/01/02. Rushton J, Lyons N. 2015. Economic impact of Bluetongue: a review of the effects on production. Vet Ital. Oct-Dec;51:401-406. Epub 2016/01/08. Zanella G, Durand B, Sellal E, Breard E, Sailleau C, Zientara S, Batten CA, Mathevet P, Audeval C. 2012. Bluetongue virus serotype 8: abortion and transplacental transmission in cattle in the Burgundy region, France, 2008-2009. Theriogenology. Jan 1;77:65-72. Epub 2011/08/30.

Keywords: Bluetongue, machine learning, Spatial Epidemiology, Europe, Risk-based surveillance

Conference: GeoVet 2019. Novel spatio-temporal approaches in the era of Big Data, Davis, United States, 8 Oct - 10 Oct, 2019.

Presentation Type: Regular oral presentation

Topic: Spatio-temporal surveillance and modeling approaches

Citation: Alkhamis MA, Fountain-Jones NM, Perez AM and Sanchez-Vizcaino JM (2019). Spatial Machine Learning Approach for Risk-Based Surveillance of Bluetongue Virus Serotypes in Europe. Front. Vet. Sci. Conference Abstract: GeoVet 2019. Novel spatio-temporal approaches in the era of Big Data. doi: 10.3389/conf.fvets.2019.05.00057

Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.

The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.

Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.

For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.

Received: 29 May 2019; Published Online: 27 Sep 2019.

* Correspondence: Dr. Moh A Alkhamis, Kuwait University, Kuwait City, Kuwait, maalkamees@ucdavis.edu