Rodent reservoirs of future zoonotic diseases

Abstract

The increasing frequency of zoonotic disease events underscores a need to develop forecasting tools toward a more preemptive approach to outbreak investigation. We apply machine learning to data describing the traits and zoonotic pathogen diversity of the most speciose group of mammals, the rodents, which also comprise a disproportionate number of zoonotic disease reservoirs. Our models predict reservoir status in this group with over 90% accuracy, identifying species with high probabilities of harboring undiscovered zoonotic pathogens based on trait profiles that may serve as rules of thumb to distinguish reservoirs from nonreservoir species. Key predictors of zoonotic reservoirs include biogeographical properties, such as range size, as well as intrinsic host traits associated with lifetime reproductive output. Predicted hotspots of novel rodent reservoir diversity occur in the Middle East and Central Asia and the Midwestern United States.

Keywords: disease forecasting; generalized boosted regression trees; machine learning; pace-of-life hypothesis; prediction.

Fig. 1.

A bubble plot showing the types of pathogens and parasites (top axis) recorded to infect rodent species in the wild. Of the rodents that harbor one zoonosis (row 1; 138 species), the majority of the etiologic agents are viruses (n = 57), and the minority are fungi (n = 4). However, among reservoirs carrying numerous zoonoses, the distribution of viral, bacterial, helminth, and protozoan etiologic agents is more even. Fungal diseases are underrepresented overall.

Fig. 2.

A map showing global hotspots of (A) rodent reservoir diversity and (B) novel reservoir species predicted by our models to be in the 90th percentile probability of harboring one or more undiscovered zoonoses (58 species). Warmer colors are the overlapping geographical ranges of multiple species, and these areas are magnified in Insets to show that hotspots occur in the Midwestern region of the United States (Kansas and Nebraska) and across the Middle East and Central Asia (Kazakhstan and northern China). B also outlines (black and maroon) the geographic ranges of three species with the highest probability (∼70%) of being undiscovered zoonotic reservoirs.

Fig. 3.

A map showing global hotspots of (A) current rodent hyperreservoir diversity and (B) 159 new hyperreservoirs predicted to be in the 90th percentile probability of harboring additional zoonoses beyond the single disease that they are currently confirmed to carry. B also outlines (black and maroon) the geographic ranges of three species with the highest probabilities (62–69%) of being zoonotic hyperreservoirs.

Fig. 4.

Marginal plots of the top 15 predictor variables from a generalized boosted regression analysis on the number of zoonoses carried by rodent reservoirs showing the marginal effect of each trait (shown in order of importance) on the probability of harboring one or more zoonotic pathogens. In general, trait values associated with fast life history strategies have the strongest influence on whether the model correctly predicts the number of zoonoses carried by reservoirs. In addition to traits compiled from the PanTHERIA database (14), three derived variables were important: relative sexual maturity age [age at sexual maturity/maximum longevity (days)], relative age at first birth [age at first birth/maximum longevity (days)], and production (35) (mean mass of offspring produced per year normalized by adult body size). The definitions for all variables can be found in Dataset S1.