To understand how the brain works we need a wiring diagram of it. If we ever get there, this diagram will give us hints about how information is processed, synthesized and perhaps even stored, and eventually will bring us closer to understanding things like memory, emotion and consciousness. So whole brain connectivity maps – aka connectomes – are a hot pursuit right now. We are just on the cusp of starting to have the capability to reconstruct wiring diagrams of neural circuitry on a relatively large scale (read: volumes of a fraction of a cubic millimetre). Knowing that replicating a very exact wiring diagram of a single brain is currently nearly impossible, some groups are opting to generate models of connectivity, which involves taking small pieces of information and extrapolating from them to make a best guess of what the connectivity likely looks like.
How best to construct a model like this? Chiang and friends in Taiwan decided to look at fruit fly brains, take heaps of pictures of individual neurons inside those brains, throw the photos together at the end and see what they got. They chose the fruit fly (Drosophila melanogaster) because the fly is a longstanding and extremely useful model organism. The fruit fly is genetically very malleable and as a result there are many techniques around that allow experimenters to probe their cellular biology. Using one of these genetic techniques unique to flies, the group forced small ensembles of neurons or even single neurons to make fluorescent proteins. They could then remove the tiny brain, take a picture of the neuron(s) and determine their 3D position relative to a structure at the centre point of the brain. Using the resulting 3D coordinates, and a complex algorithm that is beyond my comprehension and apparently beyond the scope of the paper, the 3D images of neurons were transformed into the space of a “template brain,” which is just a model based on the average of many brains. By labelling different populations of neurons in different brains with their fluorescence tool, Chiang and friends generated a bank of 3D neurons that they then populated their average brain with and they ended up getting something (a model) that looks like this:
Noteworthy developments in the paper include the group’s working definition of a “Local Processing Unit” – an ensemble of neurons that is presumably making a concerted effort in some kind of neural computation. This definition and their algorithm for pulling Local Processing Units out of their model indicates that an obscure group of neurons called the ventrolateral protocerebrum (VLP) may in fact contain, not one, but two separate Local Processing Units! Not the big conclusion you expect from a Cell Press mag. What’s more, this is just a prediction, actually testing it is a whole other kettle of fish, and since we apparently don’t know much about the VLP it will be even harder. Something a bit more exciting is that the final wiring diagram that the group constructs may be able to predict the direction of information flow. They do give a single example of this (without context), but other than that they are pretty close-chested with their predictions of info flow. All in all, the findings and non-findings of the paper amount to rather mundane predictions. Which leads to an important point: the study doesn’t really show us anything new about the fly brain. While their process is interesting, the images are beautiful and the wiring diagrams they spit out at the end are proof of a tonne of work, that impressiveness doesn’t really give rise to tangible findings.
For me, the true value of this study shines through in the transparency and scientific goodwill apparent in the work. From the beginning of the article Chiang and friends admit openly that their approach doesn’t extend past generating a map of “brain-wide interregoinal connectivity.” Indeed, their process can’t give any real insights into fine neural circuitry because they do not consider synapses. But the really impressive part is this: “Raw data for each of 16,000 sample images and the two template brains can be freely downloaded from the FlyCircuit website for offline analysis and evaluation of registration precision.” If you go here and click on the image, you can even watch a slide show of all 16,226 individual neurons. So not only can you theoretically go and check that their assertions about their work are true, you can use the data they have generated in your own analysis. You can even add your own data to the database (You being drosophilists).
Chiang and friends’ work stands as an encouraging advance in the neuroscientific community. Not only do they emphasize that the continued mapping of the fly brain is set up to be a worldwide collaborative effort, but, judging by my meager travels in science, they also set the standard for transparency in an increasingly secretive scientific institution. Perhaps we should all take example.