Brain-map.org

I read an article this week that shows how my Dad's world of manufacturing, mine of brain imaging, and computer science Google style are teaming up on neuroscience lately.

The Allen Institute for Brain Science in Seattle was started by Paul Allen, co-founded of Microsoft and the 44th richest person on earth, in 2002. The project's goal is to map gene expression for the entire human brain. Where in the brain are each of some 22,000 genes expressed? Gene expression is being visualized by a lab-bio method called "in situ hybridization"-- dripping onto slices of brain RNAs that will bind to corresponding base-pairs of DNA, then dyes that bind to the RNAs and stain them purple, "the color of spilled wine", with the shades' intensity corresponding to the concentration of a particular gene. The science is all doable. The difficulty of the Allen Institute's project was the sheer scale of it: Each individual brain would need to be sliced into thousands of one-micron-thick slices, and stained for 20,000 different genes, then each stained slice would need to be photographed and analyzed. This would take human scientists years and years of repetitive labor. The answer?

Industrialized science by robots! I've always believed that science ought to be imagined by scientists, the theory and experimental design thought out, and then all the math and data-crunching should be left to teams of unthinking drones, either robots, mole-rats, ants, or number-grubbing physicists, so a scientist's work can stick to the interesting, imaginative stuff. Now my dream of science seems to have come true. "What the institute needed was someone who could translate this epic ambition into an efficient production process, in which thousands of brain slices would be collected and assessed every day." The guy the head of the lab, Allan Jones, found to solve this engineering problem has a background alot like my Dad's: "this led Jones to hire Paul Wohnoutka, a former Boeing engineer with decades of experience in managing complex manufacturing systems ("I thought a commercial airliner was the most challenging thing I'd help build," he says, "I was wrong.")"


So Wohnoutka designed this elaborate system of bar-coded brain-slice-slides, that neuroscientists can scan with computerized gadgets to get all the data they need about the slice instantaneously. Amazingly, the slices are even analyzed by robots(!): robotic Leica microscopes with glass-slide loaders, barcode readers, and small computers running image-analysis software (that is, statistical pattern recognition algorithms like the ones we use in our fMRI work at Princeton and in Kyoto, and in Google searches). Each slice of brain converts instantly into its location in the digital brain-map, showing the expression of each of 22,000 genes in each voxel.

Here's where the project converges with the brain-mapping work I've done. The Allen Institute is the first "wet lab" project I've heard of that thinks of the brain in terms of "voxels"-- those 3mmX3mm cubes of cortex that we break the brain down into in fMRI (~40,000 per brain). The project needed to decide on a level of resolution larger than individual neurons but smaller than "brain parts", and since they don't have any over-aching theory or hypothesis about the functional architecture of brain gene expression, they just decided to think of the brain as a volume of cubes, as we do. So, they'll end up converting the brain into a "3D matrix" form a lot like the one we think about in fMRI. Anyone doing an fMRI experiment will be able to spot an active voxel and say, "I wonder what genes are expressed in that cube?" and on brain-map.org-- The FREE, open-source gene-brain map from the Allen Insitute, which will be finished by the robots in 2012-- they'll be able to see, instantaneously.

Pretty wild stuff, eh? If you're interested in this convergence of comp sci and brain-mapping, a friend of mine David Weiss explains the idea really well here.

David Weiss '07 on his Senior Thesis, Play video (3 min. 15 sec.) »
David Weiss '07, currently pursuing a PhD in computer science at the University of Pennsylvania, speaks about his Princeton thesis, for which he developed computer science methods to help neuroscientists analyze fMRI data.

David was the other undergrad advisee in Ken Norman's lab with me senior year (We were both research assistants in Ken's lab the same summer, so we got to know each other pretty well, and Dave helped with my Spidey experiment. He was a computer science major, not psych, bio or neuro, and his job was more Googley, developing the algorithsms for "dimensionality reduction"-- how to reduce fMRI activity patterns to relevent features that a pattern classification algorithm can actually parse and make sense of.) Anyway, check out this little 3 minute clip and let me know your thoughts. That computer behind Dave with the brains on it is in the scanner room I used to scan my nine friends' brains in This is Your Brain on Spiderman.

One day maybe we'll all end up teamed-up on a Google-Meets-Kyoto-meets Brooklyn indie rock- meets Chicago comedy- meets jazz project of interstellar brain decoding, where all you'll have to do is flip a switch to churn bebop, jokes or novels out of your skull. What do you say?