For the upcoming summer scholarships from AMSI, I have several projects I am proposing:
- the role of concurrent relationships in disease spread
Mathematical modeling has played a significant role in our ability to combat infect disease spread, both by helping us to choose optimal interventions and by helping to understand the role of various processes in transmission at the population scale. A mechanism which is hypothesized to play a large role in the HIV epidemic is the existence of concurrent relationships: that is, some individuals have multiple partnerships that overlap in time. This can increase the spread of disease both by allowing for more routes of transmission, but also by speeding up transmission chains as otherwise onward transmission would have to wait for partnerships to end and reform. Although these effects are clear, their magnitude is uncertain. This is in large part because mathematical techniques to predict their impact have been relatively poor. Recently developed mathematical methods allow us to more carefully analyze the role of concurrent relationships.
This project will use these recent methods to analyze the spread of disease through populations satisfying various assumptions. Some populations will exhibit pure serial monogomy, others will exhibit pure concurrent relationships, while others will interpolate between these. We will investigate the size of epidemics in these different populations and explicitly measure the roles of different individuals in transmission chains. The results will give insight into under which conditions concurrent relationships may or may not play a significant role in disease spread.
- What do we need to know to respond to the next emerging disease?
Recent years have seen the emergence of a new influenza strain, SARS, Ebola, MERS as well as other less-publicized diseases. When new diseases emerge, decisions must be made about the public health response. These decisions have significant societal and economic implications. An excessive response may worse than no response at all, but the data on which policy makers base their decisions may not be enough to make good choices, particularly early in an epidemic. Often, as a result of a focus on gathering data required for an early decision, the data needed to inform a later decision has not been collected, and so we could see a cascade of poor decisions due to inefficient data collection.
This project will use mathematical models of different scenarios to explore what data is needed to make good decisions and how it might be collected at early stages of an epidemic.
- The spread of dynamic processes on random graphs with spatial structure
Many dynamic processes happen on networks such as the spread of diseases or ideas. Some of these are well understood in random networks. However, real networks have spatial structure to them, which influences the spread. In this project we will look at a (still unpublished) class of random networks that have spatial structure. We will simulate the spread of these processes and then study the resulting outcomes, comparing them with random networks.
This has application to understanding how diseases spread, understanding how the Arab Spring spread, or any of a number of other interesting social or biological problems.
- Evidence of systematic delays by academic publishers to circumvent funding agency regulations?
A number of funding organizations (NIH, Gates Foundation, Wellcome Trust and others) have recently introduced requirements that all publications resulting from this research be freely available to the public. Publishers have been forced to accept this condition because otherwise they will not be able to publish some of the top research. However, there is a potential loophole in many of these requirements: the date at which the papers must be freely available is based on the official publication date. Some publishers put research articles online once they are accepted, but not yet officially published in paper form. They are able to charge for access to these articles during this window (one of my papers was published online a full year before it was published in print).
This project proposes to investigate whether there is evidence that publishers are taking advantage of this loophole by investigating whether there is a systematic delay in publication for these articles compared to articles which are not subject to this criterion. The project will require data gathering using automated methods (likely through Python), and the application of methods of biostatistics to measure whether there is a measurable effect.
I have a number of potential projects (and some funding) available for students or postdocs. If you are interested please contact me (and if you're faculty at another university and have time/students on your hands and an interest in some of these, please contact me as well).
- Concurrent relationships and HIV: We can derive relevant equations to allow us to analytically study the role of concurrent relationships in the HIV epidemic. The goal of this project is to derive these equations, code them, and study relevant parts of parameter space. I expect to see some interesting bifurcations, including bistability of equilibria. Small projects from this would be appropriate for an undergraduate research project, while larger scale projects could be several PhD/postdoc projects.
- Spread of interacting diseases in a network: It is possible to write down equations predicting the spread of two interacting SIR diseases. These diseases may enhance or inhibit transmission of one another. As a practical application, this can help determine the interaction between diseases such as Herpes and HIV where there is a common network component to the transmission chains of both diseases. Separating out how much of the observed overlap of these diseases is due to a common cause (the network) as opposed to a biological interaction of the two diseases would give insight into whether treating Herpes could be an effective intervention against HIV.
- Spread of antivaccine sentiment and disease: A small proportion of people refuse to vaccinate their children against vaccine preventable diseases. These beliefs are spread from person to person. This "infectious" process is more complex than a disease as it generally takes more than one exposure to the idea to convince people to refuse vaccine and sometimes being told that vaccination is safe results in the believing it is not. Modeling such a process is more complex. Many of the contacts that spread this belief also have the potential to spread the disease. This is likely to result in complex dynamics. What are the implications?
- Non-disease projects: I'm interested in some non-disease related projects, but I have not had time to get into these in depth.
- Reducing gridlock with automatic cars: Much of rush hour traffic can be traced to the fact that when a car hits its breaks, the car behind has a delayed response, and so it must slow down more, resulting in the car behind it slowing down even more... A chain reaction proceeds leading to a jamming transition. A single car allowing a larger gap to form ahead of it can break this reaction, thus speeding up the average car behind it. See this page. This comes at a cost that the car implementing this strategy may end up going slower. Could we use automated cars or cars with computer-assistance to reduce rush hour traffic? If so, what strategies should they use to drive?
- Efficient Product Rankings: Many companies use user feedback to rank the quality of products they sell. This is provided to customers for guidance. User rankings are often 1 or 5 stars, with little in between. This makes it difficult to easily distinguish between two good products, and it takes a long time before we can make reasonable comparisons between new and established products. These rankings are also susceptible to the manufacturers giving fake reviews. I have identified some methods to reduce these problems. This project will analyze various aspects of how these algorithms would perform.
- Understanding course prerequisite networks: Many different universities have their own course structure and prerequisite requirements. What can we learn about the university by studying this network, and what can we learn about the courses by studying common features of different networks? Algorithms such as community detection should help us find interesting connections between subjects, while algorithms such as PageRank could help us find which courses are most important. For practical application, this could help us identify gaps in coursework at a university.