AI Predicting Cholera From Gut Bacteria

AI technologies have become increasingly potent in a range of healthcare use cases in recent years, but undoubtedly the most common has been trawling through large quantities of data to find patterns that indicate problems faster than human doctors can spot them.

A recent study by a team from Duke University highlights the latest in this line of AI-based applications.  It sees the technology utilized to spot patterns in gut bacteria that indicate risk of cholera.

“These are patterns that even the most sophisticated scientist couldn’t detect by eye,” the researchers say.  “While some people are warning about artificial intelligence leading to killer robots, we are showing the positive impact of AI in its potential to overcome disease.”

Gut feeling

The study suggests that examining gut microbes could provide valuable insights into the prevention of cholera and similar diseases.  The algorithm was able to accurately predict who would get ill from cholera.

Cholera spreads rapidly and causes millions of cases of diarrhoea each year.  Despite the huge number of cases however, it’s still not clear why some people become sick and others don’t.  Various studies have pinpointed factors such as age and blood type, but these provide incomplete insight into the differences in clinical outcomes.

The researchers turned to AI to explore whether the trillions of bacteria that are resident in our gut might provide shed some light on the matter.

Swabs were collected from Dhaka residents who had lived in the same house as a person who had been hospitalized with cholera, and were therefore at high risk of developing it themselves.  Indeed, of the 76 households included in the study, roughly a third developed cholera during the follow-up period.

The algorithm was trained on the micribiota data from each swab alongside the results of some 4,000 different bacterial taxa from the samples.  After looking for patterns in those who got sick that distinguished them from those that did not, the algorithm identified 100 microbes that seemed responsible.

“Normally, you have to eyeball the data, studying one bacterial species at a time in hopes of finding a signal that is associated with infection,” the authors say. “Machines have the ability to look at a hundred species at a time and amalgamate them into one signal.”

The algorithm proved more proficient than the models normally used for this kind of task, whilst it also suggested hypotheses to explain the patterns found.  The findings are particularly joyful for those who have long thought that the bacteria in our guts could provide a range of insights into our health and wellbeing.

“Scientists have long had a hunch that gut bacteria might affect a person’s susceptibility to diarrheal diseases, but our study is among the first to show this in a real-world setting,” the authors conclude.

Facebooktwittergoogle_plusredditpinterestlinkedinmail