Published online 29 March 2011 | Nature | doi:10.1038/news.2011.190

News: Q&A

Allen Institute aims to crack neural code

New science chief envisions a neural 'observatory' to determine the part brain cells play in behaviour.

Christof Koch.

Christof Koch was named chief scientific officer of the Allen Institute for Brain Science last week. Researchers at the institute — set up in Seattle, Washington, with a US$100-million donation from Microsoft co-founder Paul Allen — have spent the past eight years creating three-dimensional atlases of the mouse and human brains that combine anatomical and genomic data. Koch, a neuroscientist at the California Institute of Technology (Caltech) in Pasadena, tells Nature why the institute's next big project will be to map the neural activity, or code, underlying certain behaviours, and make those data public.

Why did you take the job?

It's an unprecedented opportunity. At the Allen Institute, you can focus a tremendous amount of money, technology and people onto a particular problem in a way you can't at a university, where you have to make sure every student has their own niche, gets their PhD and gets first authorship on a paper.

What problems will you go after?

The idea is to focus on one or two behaviours — how we see, for instance, or smell, or remember — and ask how the relevant information is encoded, represented and transformed to give rise to behaviour.

The challenge is a bit like creating the Thirty Meter Telescope, which is going to be built on top of Mauna Kea, Hawaii, in the next decade, at a cost of roughly $1 billion. There you have a couple of hundred people who are all working toward a common goal.

Neuroscience hasn't had something like that, but the time is right to bring all these resources to bear onto a single question, not 20 questions in 10 different animals, each behaving differently. You essentially build a brain observatory where you try to study one behaviour exhaustively across the brain, and you make the data available to other people.

Many labs are already studying neural coding. How will your efforts be different?

Every one of these labs is using different techniques, different animals, different behaviours, electrodes, genes, statistical standards and mathematical analysis. There are basic neuroscience questions that have been around for 20, 30, 40 years that haven't been solved, because we can't agree on standards. People are more likely to use the same brand of toothbrush than the same type of microelectrodes or statistical techniques.

Is this a big shift for the Allen Institute, which is best known for its brain atlases?

It's still about the product: data that will be put out there. But it's not only data, it's understanding — what are the deep principles of neural coding? This relates to challenges of information processing and coding because that's what the neuroscience community and Paul Allen are interested in.

How will you decide what behaviours to tackle?

There will be workshops and conferences where we invite senior scientists in neural coding, modelling, electric recording and optogenetics. An advisory committee will help us decide what technologies to use, which technologies to develop and which animal species to study.


It's a challenging problem, but I think neuroscience is getting quite mature. Until recently we could only do what astronomy does: observe. But new optogenetic tools allow us to intervene, to turn neurons on and off. So we can now move from mere correlation and observation to causation. It's an exciting time.

Will you still have your own research lab?

I'll spend 75% of my time at the Allen Institute and 25% at Caltech. We're doing all sorts of very exciting stuff in my lab right now, and I don't want to leave it. We're recording the activity of single cells from human brains, we're studying the biophysics of the brain, and we're studying the neural basis of consciousness.

Will your two jobs overlap?

I've been studying consciousness for 25 years now and we've used every technique there is in volunteers and patients. But to understand consciousness, we need to be able to image the activity of millions of individual neurons at the same time. There's no way for the foreseeable 20 or 40 years that's going to happen in humans. But in mice this is feasible. This we can do in the next 5 to 10 years. 

Commenting is now closed.