Computational Ethology

New Technology is Enabling new Science

Cameras sensors have more pixels and they’ve gotten cheaper. That means we can record larger areas with high resolution video. At the same time, computer vision has been completely revolutionised by deep learning so that tracking problems that would have been difficult or impossible to solve only a few years ago are tractable. The combination of these two trends means we’ve got more quantitative behaviour data than we’ve ever had before and it’s growing quickly. The question, then, is what are we going to do with it? One answer is Computational Ethology, which is using these new tools and new data to revisit longstanding questions in the study of animal behaviour. What is behaviour, anyway? How should we represent it? Are there general principles that can guide us? Are there laws of behaviour that we might discover? We’ve written about one perspective here, but it’s a thriving field with lots of questions and few clear answers so far.

To treat symptoms you first have to identify them

One application of computational ethology that we’re particularly excited about is in identifying phenotypes in worm disease models. With genome editing, it is possible to introduce specific mutations that are known to cause diseases in humans into the corresponding worm genes. If these mutations lead to observable changes in the worms, then these become symptoms of the worm version of that disease. One of the challenges of this approach is that it’s not always possible to see the resulting difference. But by recording high resolution videos and analysing them in the right way, we’ve shown that you can detect even subtle phenotypes that can be difficult to detect by eye. We’ve continued to improve methods for tracking worms and phenotyping them and we’re actively applying these to worm disease models.

With the right phenotypes, we can then look for drugs and natural products that bring the phenotype back towards the healthy state. You can learn more about our technology for phenotypic screening or, for a more general perspective on this approach, check out the original paper on the systematic discovery of nonobvious human disease models through orthologous phenotypes.

.