At the beginning of October the world took a collective sigh of relief. In the month’s first week there were no reported cases of Ebola, something which hadn’t happened since its outbreak in March 2014.
As the media coverage died down, those of us who don’t live in the areas where the disease has caused such devastation could have been forgiven for thinking Ebola had gone away. That, however, would be wishful thinking. As recent as October 28, three new cases of the disease were recorded, with Reuters stating that the epidemic could drag into 2016.
The attempts to contain Ebola have been extensive, valiant and ultimately crucial in saving many lives. But many questions remain about how the Ebola outbreak actually began. The chain of transmission starts with Emile, a two-year old boy who died two days after the disease’s onset. His three-year old sister, Philomena, died a week later. From there, the World Health Organization was able to construct a rough chain, which you can view here.
The same question remains. How did Emile - or patient zero - contract the disease in the first place? Scientists determined that the likely cause were bats. A colony lived near the trees where Emilie used to play with his friends. That is still a theory though. The trees had caught fire before the scientists had a chance to investigate them.
Public health officials had no choice but to react to these events and potentially destroy evidence. According to Barbara Han, Disease Ecologist at the Cary Institute of Ecosystem Studies, however, this isn’t the case anymore. Machine learning and computer modeling can help predict which species have the potential to cause future epidemics.
After the initial outbreak, health officials often rush to reservoirs to see if they can determine how and where it’s been infected. Han’s use of machine learning has allowed her to create what she calls ‘caricatures’ of infected reservoirs. She states: ‘[the caricatures reveal] the suite of features that distinguish the unusual species that can harbor microbes dangerous to humans. I then use algorithms to sort through hundreds or thousands of species that have never been checked for zoonotic diseases, and calculate the probability that any given species is a disease reservoir based on its similarity to that caricature.’
If this data-driven approach is successful, it could save thousands of lives. Infectious diseases are on the rise, and many are zoonotic. Therefore, health officials are crying out for a method which allows them to be preventative, not reactive. Han is confident that Machine Learning can do that. ‘One day, I hope that biologists will forecast disease outbreaks in the same way meteorologists forecast the weather.’
The main advantage of Machine Learning is its ability to deal with complexities. With a number of variables interacting at one time, findings can become difficult to interpret. Machine Learning side steps this. On this issue Han says: ‘The algorithm doesn’t care how the variables are interacting; its only goal is to maximize predictive performance. Then we human scientists can step up.’
If only it were as simple as analyzing reservoirs. Human beings - mainly through hunting and urbanization - will always come into contact with wild animals, and therefore potentially wild diseases. This makes the prospect of predicting outbreaks, such as Ebola, even more challenging. Han, however, remains optimistic: ‘I see it as part of an even greater challenge: figuring out how to live harmoniously with the wild creatures with which we share this planet’.