Transmission networks and epidemiology

Network models have become fundamental to epidemiology. They allow powerful and flexible ways of mapping how a disease moves through a population. Such knowledge is essential to monitoring and control of disease spread. Our group has contributed to such understanding with both applied and theoretical investigations.

Tracking HIV-1 transmission patterns on an epidemic scale remains socially relevant as the number of newly diagnosed infections remains steady. Since the virus evolves at the same rate as it transmits, viral genetic sequence data leaves hints of the transmission pathway. We conducted a study based on 21844 HIV-1 subtype B gene sequences collected by the EuResist consortium. A transmission network was built by thresholding evolutionary distance between sequences. 4775 sequences showed high genetic similarity to at least one other sequence in the data. Almost always, two similar sequences originated from the same country. This suggests that HIV-1 transmission networks in Europe are highly country-specific. Our result, the largest pan-European study to date, agrees with the results of country-specific studies focusing on Germany, Italy, and Switzerland.

An epidemiological approach also let us answer an important open question in network theory. Early work in network science gave methods which identify a network's most important nodes based on network topology. For example, the node with the highest degree (number of connections) is often the most important. The last decade, however, has seen growing scientific awareness that such measures do not generalize to the rest of the network, that is the 99.9% of nodes which are not highly influential.

We measure node influence based on the expected value of the number of infection routes leading out from the local neighborhood of a node. The measure is called the expected force, since it represents the expected force of infection that would be generated by an epidemic process originating from a given node^[1]. Extensive tests show the expected force predicts epidemic outcomes with high accuracy over a broad range of network structures and spreading processes. Our most recent work evaluates its predictive power for pandemic spread over the world airline network^[2]. This study used highly realistic models which incorporate local population dynamics as well as travel patterns when simulating how a local disease outbreak would spread globally. The expected force of the airport servicing the outbreak's starting location showed 90% correlation with the virulence the disease needed to become pandemic and 87% correlation with the speed at which the outbreak achieved pandemic status.

In a world increasingly defined in terms of networks, better knowledge of network models of disease are essential to ensuring public health.

Figure 1 // Europe-wide HIV transmission network: a transmission network constructed by means of selected limit values for the evolutionary distance. The nodal color shows the country of origin of the infected patients.

[1] Lawyer, G. "Understanding the spreading power of all nodes in a network: a continuous-time perspective" Sci Rep 2015 volume 5 page 8665. doi=10.1038/srep08665

[2] Lawyer, G. "Measuring the potential of individual airports for pandemic spread over the world airline network" BMC Infectious Diseases 2016 volume 16 page 70. doi=10.1186/s12879-016-1350-4