BRAIN MAPPING

THREE BIG DOUBTS ABOUT BRAIN-MAPPING EFFORTS

Neuroscience research has come a hell of a long way since the days of scalpels and electrodes.

While some research teams are exploring the molecular machinery that churns at the hearts of nerve cells, others are working to assemble wiring diagrams for whole regions of the human brain. Just as biological science never looked the same once Watson and Crick explained the structure of DNA, neuroscience is transforming into a field filled with laser-controlled neurons, programmable stem cells and micro-scale brain scans.

Beyond all this excitement, though, looms a far more vast and ambitious goal – one whose scale and complexity exceed even the mapping of the human genome. Over the past several years, a growing group of scientists have been fighting for the idea that we can (and should) produce, within our lifetimes, a digital map of every function of every one of the trillions of synaptic connections in a human brain: A complete human connectome. Teams around the world, such as the minds behind the Human Connectome Project, are already working hard toward this goal, often freely sharing the data they discover along the way.

The Human Connectome Project’s first huge data sets are already freely available to scientists around the world.

Meanwhile, this February, the White Houseannounced the launch of the Brain Initiative, a decade-spanning effort to build a “Brain Activity Map” or BAM – a simulation, in other words, of all the activity in a human brain, from the cellular level on up. The project’s launch budget is $100 million, and some scientists expect that costs will soar into the billions before it starts cranking out useful data.

Unsurprisingly, this has stirred up a hurricane of press coverage – not all of it positive. While some advocates of the BAM project promise that it’ll unleash a wealth of new cures for neurological and psychological diseases, opponents argue that even billions of dollars and years of research won’t be enough to decode the brain’s workings on such a comprehensive scale – especially if, as some anti-BAM pundits say, we’re still a long way from knowing how the brain even encodes information at all.

I’ve put together a little write-up on three of the biggest BAM bones of contention. Though I can’t cover the whole issue in detail with just one article, these summaries should help you score some points in a BAM-related argument – and give you some fuel for your own exploration. So let’s see what (some of) this fuss is all about.

Doubt #1: Do we have the computing power to simulate a whole human brain?

Nvidia’s “Titan” supercomputer, which (as of April 2013) holds the world speed record of 20 petaflops.

The BAM invites a lot of comparisons – both positive and negative – with the Human Genome Project of the 1990s. Both are long-term projects, both are hugely expensive, and both involve number-crunching and analysis on scales that demand tight cooperation from top scientists and universities around the globe.

But whereas the Human Genome Project set out to map somewhere in the neighborhood of 20,000 to 25,000 genes, all of them constructed from the same four nucleotide molecules, a map of the human connectome would have to incorporate the behavior of at least 84 billion neurons and as many as 150 trillion synapses – all communicating via a dizzying menagerie of messenger chemicals, not to mention physically reshaping themselves as a brain grows and learns.

Estimates vary widely on the question of how much computing power it’ll take to simulate a whole human brain, but even the most optimistic experts believe it’ll take a computer capable of performing at least 1 quintillion (that’s 1,000,000,000,000,000,000) floating point operations per second (1 exaflop). By comparison, your average home computer processor maxes out around 7 million flops (7 gigaflops), a fast graphics card can reach over 300 million flops (300 gigaflops), and the latest supercomputer processors clock in at a little over 20 quadrillion flops (20 petaflops). So, on that front at least, our resources are rapidly approaching the goal – scientists at Intel predict that we’ll be computing in exaflops before this decade is over.

But raw computing power is only one part of the equation. In the most basic sense, even the most advanced computer is just a machine that follows instructions – so even after we’ve built our exaflopping supercomputer, we’ll still need to know what instructions to give it.

Carlos Brody, a neuroscientist at Princeton’s Brodylab, has added a clarification of his own to this section. Here’s what he has to say:

“I think Doubt #1 is about the European Human Brain project, not about the U.S.-based BRAIN Initiative. The way I’ve understood it, the Europeans, with their billion-euro Human Brain project, are trying to simulate every neuron in a brain. In contrast, the U.S.-based BRAIN Initiative/BAM is about developing the technology to allow us to record the activity of every neuron in a brain. Not simulate, but measure what’s there. It’s a big difference, because in order to simulate you have to build in a lot of knowledge we don’t yet have (i.e., put in a giant truckload of untested assumptions). That is largely why many people think the simulation effort is pointless, there’s so many untested assumptions going in that what you end up with may bear little to no relation to an actual brain. The goal of measuring the activity, as in BAM, is to gain that knowledge we don’t yet have.”

Thanks, Dr. Brody, for your insight into that distinction!

Doubt #2: Do we know enough about brains to know what we’re attempting?

The Human Genome Project set out to map the position – but not necessarily the function – of each nucleotide in all 23 human chromosomes.

Contrary to oft-repeated belief, the Human Genome Project’s goal was never to decode the function of every gene in human DNA – it was to map (sequence) the order and position of every nucleotide molecule in all 23 human chromosomes.

Scientists have only begun to make a dent in decoding the 20,000+ genes whose positions the Human Genome Project mapped. Even today, leading researchers are still debating how many genes the human genome actually contains – let alone what functions most of those genes encode. And that’s more than half a century after Watson and Crick described, in detail, the way that DNA encodes recipes for manufacturing the molecules that make up our bodies.

When it comes to the brain, on the other hand, the world’s top neuroscientists are still puzzling over the question of how neural activity encodes information at all. We’re using computers to construct videos of entire visual scenes based on the brain activity of people watching them – but that’s only after recording brain scans of dozens of patients as they watched hundreds of videos, then telling a computer to reverse the process and assemble a video that matches the brain activity patterns it sees.

This is no small achievement, to be sure – but even so, it’s sorta like learning to recognize whether the letters in a book are Chinese, Japanese or Arabic (assuming you don’t read any of those languages). You might be able to match a new book with the country that produced it, and maybe even recognize whether it’s, say, a novel or a dictionary. But none of that tells you much of anything about what a specific line on the page actually says.

This is one of the trickiest questions for BAM advocates to answer – and the answers tend to come in two main flavors. One response is that the fastest way to crack the neural code is to try simulating it digitally – just as the fastest way to learn a new language is to start writing and speaking it yourself. Another response is that a base-level understanding of the code may not be necessary for a rich and detailed understanding of how a brain works. Scientists have already mapped the functions and interactions of all 302 neurons in the nervous system of the tiny roundworm known as C. elegans. Even without knowing exactly how these neurons encode information, we’ve still built up a precise understanding of how each of them influences other neurons and muscle cells throughout the worm’s body.

Although the human brain’s 84 billion neurons aren’t exactly a small step up from C. elegans‘s 302, it stands to reason that if we do develop software and hardware that can simulate all our neurons’ interactions, we’ll be in a much better position to pinpoint specific processes and problems down at the cellular level.

Doubt #3: Will an epic mapping project produce useful results?

Just as no two humans share exactly the same set of genes, no two human brains are wired in exactly the same way.

BAM critics like to draw a third unflattering parallel between the BAM project and the Human Genome Project: As the Human Genome Project approached completion, its White House advocates predicted that a sequenced human genome would lead to cures for diseases like cancer and Alzheimer’s, along with “a complete transformation in therapeutic medicine.” But more than a decade after the Project’s completion, very few of those medical benefits have actually materialized.

What has resulted from the Human Genome Project is a vast storehouse of data on how human DNA differs from that of other animals – and from one human being to another. This means that when we consider the outcome of the BAM, it’s important to keep our sights not on vague and grandiose promises about cures for poorly understood problems, but on what we can be sure would come out of a successful BAM project: A more detailed, accurate and integrated understanding of the human brain’s workings than we’ve ever had before.

If one thing about the BAM is certain, it’s that the project’s news coverage – and the intensity of the debates that coverage stirs up – will increase in step with the Brain Initiative’s funding demands and timing estimates. As I said at the beginning of this article, a few thousand words aren’t nearly enough to cover all the ink that’s already been spilled in the earliest stages of this debate – so jump into the comments and chime in with your own opinions, doubts, speculations and questions. Because in the end, the only way to resolve an argument is to talk it out.