Using Complex Inference Networks to identify and understand the ecological interactions in zoonoses
The latest Parasitology Paper of the Month is “Understanding transmissibility patterns of Chagas disease through complex vector–host networks” by Laura Rengifo-Correa, Christopher R. Stephens, Juan J. Morrone, Juan Luis Téllez-Rendón and Constantino González-Salazar.
Parasitic diseases are extremely complex, associated with a very large number of risk factors; from the microscopic, such as genetic susceptibilities, to the macroscopic, from behavioural characteristics of their hosts through to large scale ecological conditions. Vector borne diseases are particularly complex, as besides from purely human risk factors, they can include both non-human hosts and vectors. Unfortunately, an empirical characterisation of the impact and relevance of all possible risk factors is not feasible. There are just too many. Even restricting to just the host range of a zoonosis, and a measurement of the competence of each host, presents enormous problems. The question then becomes: Can one better identify and understand the ecological interactions that are so important in zoonoses without having to make a direct observation?
The answer is yes, using Complex Inference Networks (CIN), as illustrated in our paper “Understanding transmissibility patterns of Chagas disease through complex vector–host networks”. Unlike other ecological networks, such as food webs, where the network links are just representations of known interactions, in CIN the links are inferential. The basic idea has two inspirations: first, the old adage – “you can judge a man by his friends”. In this case you can judge a disease host by its vector “friends”. In order to be a host of a vector-borne disease it is a necessary condition that the vector and potential host co-occur. This, of course, is not a sufficient condition. However, in the light of the large amounts of point collection data available nowadays, it is a way to leverage such data to make concrete predictions. The question to be answered is: is the degree of co-occurrence more than one would expect by chance? If so, there must be a reason.
This leads me to the second motivation for the methodology: Basically, all we know about the fundamental interactions – gravity, electromagnetism etc. – has been gleaned by observing the relative positions of bodies in space and time. As an example, Tycho Brahe observed planetary motions with precision, Johannes Kepler discovered phenomenological rules associated with those motions and, later, Newton explained those rules by positing an interaction between the bodies. Thus, gravity and its properties were deduced by observing the relative positions of massive bodies. One can then ask to what extent the existence and nature of ecological interactions can be deduced by just observing the relative positions of species, as proxied by point collection data? Although this might seem far fetched, the consistency of our results with known facts across multiple diseases and, moreover, the ability to make predictions about unknown host-vector relations that have been confirmed empirically shows that analysis of geographic relations alone can be used to infer ecological interactions.
In the present paper we restrict attention to Chagas disease, but the methodology is general, and is a way to generate hypotheses and predictions about any zoonosis or, indeed, any ecological system. The modelling methodology is available (http://species.conabio.gob.mx) and we hope the community will use it as a new research tool in emerging diseases.