Brain data needs data brains
Neuroscience data is being churned out at a phenomenal rate. But who for?
Neuroscience is drowning in data. Every year, 30,000 neuroscientists meet at the Society for Neuroscience conference for five days. They have to choose which of 16,000 posters to see, each A0 or larger, each a snapshot of someone’s current research project. Fitting all these posters requires a hall a mile long. We’ve lost some great minds in there — usually in row JJJ.
But who is all this data for? At the moment: no one. Without neuro-data people, we risk missing out on the big insights into how our brains work.
Conservative estimates suggest we number at least 100,000 neuroscientists world-wide. Each pursuing the big scientific questions of brain function — how does the retina process light? Is cortex one big general purpose computer? How do neurons learn? Why does the cerebellum need more neurons than the rest of the brain put together, if when we cut it out nothing happens at all? Do rats like carbonated beverages (1)?
But given how much data is generated by all these researchers, isn’t it just possible that the answers to these questions are already out there? That the experimental data already exists?
To know this we need two things. First, the data from experiments has to be made widely available. Access to crucial data cannot depend on knowing a guy who knows a guy. Second, there has to be teams of people with the necessary mix of neuroscience knowledge and analytical skills to test the data for the answers. We have the first, but not the second.
Neuroscience, like all biology, is an inherently forward-looking enterprise. It rewards the shiny and new, the digging up of more detail, and the ability to describe it all in a glorious riot of statistics (2). Counteracting this is a growing movement across all branches of biology to publish data — to make the data publicly available for re-use and study by anyone. In other words, to make it open access.
Systems Neuroscience has been the unruly child here, kicking and fighting against this, a reflection of the shear difficulty of obtaining good data about the activity of microscopic neurons. Labs can invest years in getting enough of this data to reach any conclusions. No surprise that sharing it is not the first thing on their minds.
Yet even here data-sharing is gaining traction. The CRCNS.org data repository is slowly but surely notching up prized datasets; the Neurodata Without Borders initiative is solving the problem of putting all this data into a common format; the Public Library of Science journals now have a compulsory mandate to share the data from any paper they publish; and the privately-funded Allen Institute publicly releases its data, to a strict timetable. Indeed, they have just released an extraordinary set of recordings from the visual cortex of mice (3).
The open-access neural activity data are now there to tackle the big questions in neuroscience. But by whom? Who are these teams of people that, up until very recently, were sitting around, sharpening their pencils, waiting for data to turn up? No one, of course. There are neuro-data; but no neuro-data analysts.
We need to build neurodata teams. We have the rich, complex data needed to answer big questions of neuroscience publicly available right now. As more funders and journals mandate public data release, the amount of data will grow exponentially. And these data can only grow in size and complexity, as neurotechnology inexorably advances. We need teams of experts who can tackle these data head-on; who can make sure these data do not sit, untouched and unloved, in some dark, dusty corner of the Internet.
Having pushed for the release of data, funders of neuroscience research must now fund the teams to analyse that data (4). The financial side stacks up. Neurodata teams are highly cost effective. No animals, no reagents, no microscopes, no electrodes — no lab at all. They just need people, some comfy chairs, and a computer or two. (Oh, and caffeine; musn’t forget the caffeine — file it under “consumables” on your grant applications).
And with just that, they have the potential to find the answers to the big scientific questions about brain function.
There are tantalizing hints of this potential being fulfilled. Existing data has been used to reveal the patterns of synaptic connections between cortical neurons; to show that activity oscillations in hippocampus encode position; and to show that prefrontal cortex might use probabilities to represent the world.
Neurodata teams could turn these glimpses of a deep understanding of the brain into a barrage. It’s worth a punt.
Want more? Follow us at The Spike
Twitter: @markdhumphries
(1) Real experiment
(2) Or so you’d think, reading a society journal neuroscience paper.
(3) Why study mouse vision, when they’re nocturnal animals? That’s a discussion for another time…
(4) The Brain Observatory project, recently announced by the USA’s National Science Foundation (NSF), is the first tangible step in this direction. They explicitly plan to fund theory teams; but, their mandate is to develop new theoretical frameworks and then test them in collaboration with experimental neuroscientists. The NSF’s funding call does not consider that these teams could use existing data.
