New models for the identification of primate species likely to spread Zika in the Americas

Lead author Barbara Han, a disease ecologist at Cary Institute, explains, "When modeling disease systems, missing data can affect our ability to predict where people are at risk. Primate species have been confirmed positive for the Zika virus, interested in how the marriage of two modeling techniques could help us overcome the limited data available on the biology and ecology of primates – for the purpose of Identify monitoring priorities. "

The recent Zika epidemic in the Americas has been one of the largest in the world, affecting more than half a million people. Like other flaviviruses transmitted by mosquitoes, Zika circulates in the wild. Primates can serve as reservoirs of diseases of contagion infections in areas where mosquitoes feed on both primates and humans.

By badyzing the data on flaviviruses and primate species known to transport them, and comparing these traits with 364 primate species present in the world, the model identified known flavivirus transporters with an accuracy of 82% and badigned risk scores to other primate species likely to carry the Zika virus. The final product includes an interactive map that takes into account the geographic areas of primates to identify hot spots where people are most at risk of contagion through Zika.

Primate species in the Americas with a Zika risk score greater than 90% included: bushy capuchin (Cebus apella), the Venezuelan red howler (Alouatta seniculus) and the white-faced capuchin (Cebus capucinus) – species adapted to the life of populations in developed areas. Also on the list: White-fronted Capuchins (Cebus albifrons), usually kept as pets and caught for trade, and spider monkeys (Saimiri boliviensis), which are hunted for bushmeat in parts of their range.

"These species are geographically widespread, with abundant populations that live near population centers, they are well-known looters and are kept as pets.The people display them in the cities as tourist attractions and hunt them for Bushmeat is an extremely alarming result, "said co-author Subho Majumdar.

Adding to the problem: the mosquito species most likely to spread Zika are usually found near humans and can thrive in natural and altered landscapes.

The model

To fill gaps in the data, the team combined two statistical tools, multiple imputation and multi-tagged Bayesian self-learning, to give primate species a score of risk indicating their positivity potential for Zika.

Pathogens

The characteristics of six mosquito-borne diseases were evaluated: yellow fever, dengue fever, Japanese encephalitis, St. Louis encephalitis, Zika virus and West Nile virus. Three of them had known primate reservoirs.

Primates

The biological and ecological characteristics of the 18 species of primates tested positive for any mosquito-borne flavivirus were compared to the characteristics of 364 primate species present in the world. 33 characteristics were badessed including metabolic rate, gestation period, litter size, and behavior. The characteristics were weighted for their importance in predicting Zika's positivity.

Han explains: "Like all pathogens, the Zika virus has unique requirements for what it needs in a host animal, and to determine which species can host Zika, we need to know what these traits are, what species they have. these traits and which of these species can transmit the pathogen to humans.This is a lot of information, much of which is unknown. "

A statistical method called multiplication and imputation coupled equations (MICE) was used to overcome data limitations. MICE defines computer algorithms that consist of performing searches in feature datasets of organisms to establish links between organisms with similar or related characteristics. When the algorithm encounters a missing data entry, it uses these connections to derive the missing information and fill in the "spaces" of the data set.

Machine learning has been applied to this "knowledgeable" dataset to predict which primate species are most likely to carry the Zika virus. The model generated a risk score for each species by combining the history of flavivirus infection and biological characteristics to predict the probability of positivity for Zika.

This method could help improve predictive models for other disease systems beyond Zika. Lead author Kush Varshney of IBM Research explains, "Data gaps are a reality, especially for infectious diseases caused by wild animal hosts. Models like the one we developed can fill some of these gaps and help identify species of concern to refine surveillance. , provide benefits and help guide the efforts of the public health community. "

Varshney added, "Making machine learning on small, incomplete and noisy data sets to support critical decision making is a common challenge across many industries and sectors." We will certainly use the experience acquired in the framework of this project in many areas of application. "

Han concludes, "This research has been made possible by innovations from the wider scientific community, using primate and pathogen data collected by hundreds of field researchers, and the basic methods of research. We have adapted to this research already existed Partners At IBM, research occupied a lion's share of mathematics and coding, an extremely fruitful interdisciplinary collaboration of the kind we need more if we want to find new solutions to complex problems. "

Explore further:
Where will the next Zika, West Nile or dengue viruses come from around the world?