One of the great challenges in modern science is to understand the structure of the human brain. In particular, neurologists want to work out the complete map of connections between neurons, it’s a wiring diagram and a structure known as the human connectome.
That’s a significant challenge. So far, researchers have successfully constructed the connectome of only one creature, the nematode worm C. elegans with the grand total of 302 neurons and 8000 connections between them. By contrast, the human cerebral cortex contains some 10^10 neurons linked by 10^14 synaptic connections.
One reason why progress has been slow is that the process of mapping a connectome is enormously time-consuming. The neurons have to be imaged using electron microscopy and the resulting images stacked and aligned so that every part of each neuron can be mapped by hand. For the C. elegans connectome, this took 50 person-years of analysis so faster mapping techniques are clearly needed.
That’s why biologists are racing to develop techniques that automate the mapping process. And there are several promising candidates that should increase data collection rates by several orders of magnitude in the coming years.
But that creates another problem. The wiring diagram and its spatial layout do not uniquely specify the function of a neural system. To gain a full understanding of what’s going on, biologists need to know the types of cells involved and how they connect to each other in microcircuits. That helps distinguish, for example, the function of circuitry involved in retinal tissue from that in brain tissue.
But this kind of analysis is only possible by studying each neuron individually and marking its place on the wiring diagram, another labour intensive and time-consuming process. What’s needed is a way to automate this process as well so that it dovetails with the vast amounts of data that automated mapping techniques are likely to produce in the next few years.
Today, Eric Jonas at the University of California, Berkeley, and Konrad Kording at Northwestern University in Chicago, say they’ve solved this problem. These guys have developed an algorithmic technique that uses the spatial wiring diagram from an organism as an input and then uses it to automatically identify cell types, the circuits they form and hence the function of these neural systems.
They’ve tested their algorithm on the connectome of C. elegans and say it reveals the cell types and interesting structures within it. They’ve tested it with equal success on the connectome of the mouse retina. And their approach even automatically discovers the ‘connectome’ of the 6502 microprocessor used in the Apple II computer.
Here’s how they tackle this problem. Anatomists have discovered numerous different types of cells in neural circuits that perform vastly different functions. For example, there are some 70 different cell types in the retina which can be broadly grouped into three: ganglion cells that transmit information to the rest of the brain, bipolar cells that connect different cells and amacrine cells that feed into the ganglion cells.
These different cell types vary in features such as the length of their cell bodies, the number of connections they make and so on. For example, cells that are 10 µm long are clearly of a different type to those that are 250 µm long.
That’s the basis on which the new algorithm distinguishes cell types. It mines the connectome data looking for how the connectivity of cells changes over distance. It then groups the cells according to their discovered properties, such as their length, number of connections they make and so on. Each group should therefore be a different cell type.
To find out how well this algorithm works, Jonas and Kording have tested it on the few connectomes that have already been mapped in detail, such as the C. elegans connectome, the mouse retina and an artificial connectome in the form of the 6502 microprocessor from the Apple II computer.
The results are impressive. “The algorithm produces a good fit to human anatomist judgments,” say Jonas and Kording. “We have found that combining the available data types allows us to discover cell types and microcircuitry that were known to exist in the systems based on decades of previous research.”
It also works on artificial systems. The algorithm identified the known architecture of the 6502 microprocessor and picked out transistors involved in the internal clock, others serving as inverters and various other artificial layouts.
There are limitations, of course. The algorithm does not identify all of the cell types that anatomists have found, although more data should help. What’s more, the automated process becomes slower as the amount of data increases. “Our algorithm required several hours for 1000 neurons,” say Jonas and Kording.
Just how this can be scaled by the many orders of magnitude required for the human connectome isn’t clear. But this is a general problem in computer science and one that numerous researchers are working to improve.
Overall, this is impressive work that has significant potential for the future. Until now, the mapping of neural microcircuitry has involved a huge amount of manual labour. And even then, there is a certain amount of disagreement over what constitutes a type of cell.
The automated gathering of this kind of data will revolutionise the science of the connectome. So it’s vital that there are algorithms that can analyse it too. As Jonas and Kording put it: “Eventually, algorithms will have to become a central tool for anatomists, as it will progressively become impossible for humans to parse the huge datasets.”
It should also allow the comparison of cell types on a much broader scale, between animals and species, for example. That will be particularly important when different cell types emerge because of the stimuli they receive rather than the biomolecular properties of the cells themselves.
For example, in the primary visual cortex there are cell types that seem to select for colour and orientation in different animals but are otherwise identical in terms of molecular properties. “This would allow comparative quantitative anatomy across animals, and aid the search for the ultimate causes of connectivity,” they say.
There’s an interesting comparison here with other areas of biology that have been revolutionised by bioinformatics. Molecular biology has become a new science thanks to gene-finding algorithms and evolutionary biology has been revolutionised by computational phylogenetics.
A similar revolution is clearly in the offing for connectomics.
Ref: arxiv.org/abs/1407.4137 : Automatic Discovery Of Cell Types And Microcircuitry From Neural Connectomics