The Wuhan coronavirus has spread globally in a short period of time, infected almost 40.000 people and killed over 800 people (as of the 9th of February). The Wuhan coronavirus can be seen as an ongoing epidemic and needs to be researched and contained with the highest priority. In response to the ongoing outbreak John Hopkins University developed an online dashboard to track the outbreak.
This dashboard can be found here.
A few weeks before the Chinese government announced a travel restriction to Wuhan, a bunch of medical AI experts were already detecting the early signs of an epidemy.
During an outbreak of this magnitude time is of essence. The earlier the warning the better the chance to contain the disease.
Bluedot, a Canadian based startup, developed an AI-driven health monitoring platform in 2014. Bluedot uses machine learning techniques to analyse global news reports, airline data and reports of animal outbreaks to predict epidemics that might happen in the near future. Epidemioligists will verify the results produced by the machine learning algorithm and if everything looks correct the company will send an alert to its clients. Bluedot predicted the outbreak of a virus with as epicentre the Wuhan region and alerted its clients on December 31st, well ahead of notifications from the World Health Organization.
Using the data that is used in the dashboard by JHU we can make a machine learning algorithm that predicts on daily basis how many infections there will be and how many deaths are expected. This triggered James Ross, co-founder of fintech startup HedgeChatter to build an AI model for estimating the global reach of the coronavirus.
The results of Ross’s model were shocking. From 50,000 infections and 1000 deaths (which is accurate as we can see this data in the current dashboard) to 2.5 billion infections and 53 million deaths after 45 days.
After just a few weeks the model predicts an exorbitant amount of infected and dead people.
Luckily the model does not know every factor and Ross acknowledges that. The conditions and data fed in this model change regularly and those conditions will change the results massively. Also, now that the Wuhan coronavirus is in every news article around the world the behaviour of the people changes as well. They become more careful in hygiene and where they travel. As a result, the risk on infection is reduced.
The AI model is correct in the predictions it made if the conditions stayed the same (No global news on the virus and no attempt to contain the virus). However in an everchanging world the predicted numbers will be much lower than predicted.
With the amount of deaths still rising it is important to find a cure for the virus as quickly as possible.
Insilico Medicine used AI to identify and make small molecule structures that can form an effective treatment against the Wuhan coronavirus. It took the AI system of Insilico 4 days to identify thousands of potential new molecules. They found these molecules by using generative adversarial networks(GANs), the same technique that is used for creating deepfakes, to create new molecules. After that a machine learning filter is applied to favour the molecules with drug like properties (after all we would like the treatment not to kill people). A last filter makes sure that the overall set of molecules is unlike already existing molecules to ensure they have a diverse set of molecules to test. Four days later the algorithm has generated hundreds of thousands of new molecule designs. The most promising molecules now need to be analysed and tested. Hopefully an effective drug for the Wuhan coronavirus will be found soon.
How do you think AI can help prevent further outbreaks of diseases? And do you think AI can help combat diseases by making strategies for treatment? Let us know in the comments!
Back to blogs