Drosophila melanogaster, the common fruit fly, is in some ways a simple creature. But in others it is so complex that, as with all life, we only scratch the surface of its understanding. But researchers have taken a major step with Drosophila by creating the most accurate digital twin – at least about how it moves and, to some extent, why.
NeuroMechFly, as the EPFL researchers call their new model, is a “morphologically realistic biomechanical model” based on careful scans and careful observation of real flies. The result is a 3D model and motion system that, when asked to do things like walk around or respond to certain basic stimuli, much like a real fly would.
To be clear, this is not a full cell-by-cell simulation, on which we have seen progress in recent years with much smaller microorganisms. He doesn’t simulate hunger, vision, or any sophisticated behavior – not even the way he flies, only the way he walks along a surface and grooms himself.
What’s so hard about it, you ask? Well, it’s one thing to get close to that type of movement or behavior, and make a little 3D fly that moves more or less like a real one. It’s quite another to do so to a precise degree in a fully physically simulated environment, including a biologically precise exoskeleton, muscles, and a fly-like neural network that controls them.
To make this very precise model, they started from a scanner of a fly, in order to create the morphologically realistic 3D mesh. Then they recorded a fly walking in very carefully controlled circumstances and very precisely tracked the movements of its legs. They then had to model exactly how these movements corresponded to “physically simulated articulated body parts, such as head, legs, wings, abdominal segments, proboscis, antennae, halteres”, the latter being a kind of motion sensing device that helps during flight.
They showed that these worked by bringing the precise movements of the observed fly into a simulation environment and playing them back with the simulated fly – with the real movements correctly mapped to those of the model. Then they showed that they could create new gaits and movements based on these, letting the fly run faster or more stably than they had observed.
Not that they enhance nature, exactly, just showing that the simulation of fly motion extended to other, more extreme examples. Their model was even robust against virtual projectiles… to some degree, as you can see in the animation above.
These case studies reinforced our confidence in the model. But what interests us most is when the simulation fails to replicate animal behavior, pointing to ways to improve the model,” said Pavan Ramdya of EPFL, who leads the group that has builds the simulator (and other Drosophila-related models). Seeing where their simulation breaks down shows where there is work to be done.
“NeuroMechFly can improve our understanding of how behaviors emerge from the interactions between complex neuromechanical systems and their physical environment,” reads the abstract of the paper published last week in Nature Methods. By better understanding how and why a fly moves the way it does, we can also better understand the systems that underpin it, yielding information in other areas (fruit flies are among the most widely used experimental animals) . And of course, if we ever wanted to create an artificial fly for some reason, we would definitely want to know how it works first.
While NeuroMechFly is in some ways a huge leap forward in the field of digital life simulation, it is still (as its creators would be the first to recognize) incredibly limited, focusing only on specific physical processes and not the many other aspects of the tiny body. and the spirit that make one Drosophila a Drosophila. You can view the code and possibly contribute on GitHub or Code Ocean.