AI Case Study
Researchers at Columbia University automate the process of sex-sorting tsetse flies with machine learning
Zelda Moran and Szabolcs Márka of Columbia University have used a machine learning algorithm to automate the process of sex-sorting tsetse flies. The system is trained on scanned images of large numbers of pupae, enabling it to be able to distinguish between male, female, unidentifiable or dead. Once the male flies are isolated and sterilised using gamma rays they can be shipped off for release. Females flies that mate with sterile males have no offspring and thus both the population of tsetse flies and the occurrence of sleeping sickness decreases.
"In 2014 Ms Moran, who was then a researcher at the entomological laboratory of the International Atomic Energy Agency, in Vienna, which does a lot of this work, noticed that female and male tsetse pupae develop differently. Adult flies emerge from their pupae 30 days after pupation. Although tsetse-fly pupal cases are opaque, Ms Moran found that in certain lighting conditions, such as infrared, it was possible to observe that the insects’ wings began to darken beforehand. In the case of females, this happens around 25-26 days after pupation. In the case of males it happens later: 27-29 days after pupation. In principle, that gives a way to sort the flies before they emerge from their pupae.
Acting on these observations had to wait until a chance meeting Ms Moran had with Szabolcs Márka, an astrophysicist at Columbia. At the time, Dr Márka was using machine learning, a form of artificial intelligence, to find patterns in large sets of astrophysical data. He suggested applying a similar process to the sexing of pupae.
First, Dr Márka and his colleagues used an infrared scanner to create images of large numbers of pupae. Then, they employed these images to train a computer algorithm to decide whether a pupa being scanned is male, female or unidentifiable. It can also tell dead pupae from living ones. That done, it is a simple matter to automatically sort the living male pupae from the rest, with a machine using puffs of air or jets of water to blow the unwanted pupae away. Thus isolated, the male flies can be irradiated and shipped off for release.
This should simplify the production of sterile males for projects such as the one in Senegal. It might work on other species, too. If sex-sorting by machine learning could be applied to insects like mosquitoes, illnesses such as malaria and dengue might also come under tighter control."
R And D
Core Research And Development
"Female tsetse flies mate for life. Or, to put it more accurately, they mate once in a lifetime. That gives those who would control these pests an opportunity. A female that mates with a sterile male will have no offspring. Arrange for enough such matings to occur and the result will be fewer tsetse flies—and, with luck, less sleeping sickness, a disease spread to people and cattle by the flies.
A project under way in Senegal suggests this works. For the past five years, male tsetse flies sterilised using gamma rays have been dropped three times a week over infested areas. This has pushed the local fly population down by 98%, with a concomitant fall in sleeping sickness. But projects like this require huge numbers of sterile males to be bred and delivered in a timely manner. And that is hard.
One problem, since breeding males necessarily involves breeding females, too, is sorting the sexes, so that only males are irradiated and released. (Simply irradiating both sexes is problematic; a higher radiation dose is required to sterilise females, for example, which risks killing or disabling the males.) To sort the tsetses means waiting until the flies emerge from their pupae, chilling them to reduce their metabolic rates and therefore their activity, and then separating males from females by hand, with a paint brush. Male flies can be identified by external claspers that make them distinguishable from clasperless females. This process is effective, but time-consuming and labour-intensive."
Local fly population has decreased by 98% resulting in a fall in sleeping sickness
"Dr Márka and his colleagues used an infrared scanner to create images of large numbers of pupae. Then, they employed these images to train a computer algorithm to decide whether a pupa being scanned is male, female or unidentifiable. It can also tell dead pupae from living ones."
Images of large numbers of pupae created with an infrared scanner