Human Connectome Project: the great controversy begins
This is one of those topics that is a *very* big deal…but for which it is going to be difficult to convey the import, owing to the technical involvement of the subject. Try to hang with me on this, though–I will do my best to make it digestible!
The necessary starting point is to understand the NIH’s push for a Human Connectome Project. The “connectome” refers to a yet-to-be-created map of the wiring connections in the human brain. This is a very big deal! I don’t think that I can overstate how essential this map will be for revolutionizing our understanding of phenomena such as human cognition and psychiatric disease. By comparison, the Human Connectome will be for neuroscience what the Human Genome is for molecular biology. And, like the Human Genome, the completion of the Human Connectome map will only be the beginning (and a darn-exciting beginning) to better unfolding the layered complexity we each carry within our skulls.
Steve Petersen at Washington University is the chair of the fMRI portion of the Human Connectome Project (hereafter “HCP”). The fMRI team has laid out a two-phase timeline for completion of the project: phase I will be a two-year period of selecting methodology; phase II will be dedicated to the acquisition and analysis of the brain imaging data. Over the five-year course of the project, brain scans will be acquired from over 1,000 participants, looking at their brains’ activity patterns during both resting states, and during engagement in cognitive or performance tasks. The overall strategy is thus: 1) to use task-related brain scans to identify functional networks in the brain, and 2) to use the resting-state data to determine the finer-scale connectivity between brain regions.
Enter Randy Buckner, stage left.
At the Society for Neuroscience meeting last month, Randy Buckner–a big player in the fMRI arena from Harvard University–had two posters from his laboratory that presented a direct challenge to the meticulous planning of the HCP investigators. The posters were entitled “Estimates of surface-based cortical networks using intrinsic functional connectivity from 1000 subjects” and “Estimates of cerebellar, thalamic and basal ganglia circuits using intrinsic functional connectivity from 1000 subjects.” The response from conference goers to the Buckner lab posters was quite dramatic. I overheard one scientist ask a colleague, “So, would you say that the Human Connectome Project just got scooped?” To which his companion replied, “No. I would say they just got their asses kicked to Tijuana.”
Randy Buckner’s lab–in contrast to the HCP plans–relied exclusively on resting-state data to create sweeping maps of connectivity within the human brain. They (quite impressively) imaged the brains of 1,000 subjects during an intense scanathon lasting only a few months. That is huge throughput of subjects.
The controversy, then, comes down to very critical questions about whether resting-state scans–that is, brain scans in which the subjects are in a neutral, wakeful state–provide the correct information by themselves about large-scale brain networks. It is an open question. (Well…it’s an open question depending on who you talk to.)
There’s a case to be made for both sides, I think. In the “resting-state-AND-task-engaged-scans” camp, a chief argument is that conclusions about the boundaries of functional networks are only meaningful when you are looking at data in which those networks are actively engaged. Just from my own research on the insula, I can attest firsthand to the fact that functional connectivity outcomes do indeed fluctuate when the subject is engaged in a cognitive task that draws upon a particular brain region, versus when the subject is only scanned during a wakeful-rest state. In the “resting-state-scans-ONLY” camp, the argument is that the markers of intrinsic network connectivity are preserved in the brain, regardless of the functional engagement of the subject; and to let the resting-state data speak for itself by using k-clustering algorithms to define candidate networks, rather than imposing network masks derived from the task-positive brain scans.
Steve Petersen will have an article in a forthcoming issue of the Annals of the New York Academy of Sciences, in which he will lay out the case for the necessary inclusion of task-related scans in order to create an accurate connectome map in humans. I look forward to reading the article, and to further understanding the case for the necessity of task scans in functional connectivity analysis.
At the end of the day, though, the final answer will be provided when the HCP completes its connectome map. At that point, team “task-AND-resting” and team “resting-ONLY” can compare notes, and hash out the explanations that underpin any differences in their results.
To further draw the parallels between the Human Connectome Project and the Human Genome Project, when the NIH announced the Human Genome Project, they announced that it would take a span of years (I believe their initial projections were 15 years?) to complete the sequencing. Then Craig Venter–a private interest maverick–entered the scene and said that he could complete the sequencing of the entire human genome in 18 months. The competition introduced by Venter created a major acceleration in the NIH sequencing project, and ultimately resulted in a joint announcement of the project’s completion by both Venter and the NIH.
As an optimistic take-home message, I think that the strong differences in opinion, and the subsequent competition that it introduces, will ultimately prove to be a strength working toward the purity of the science, even though it will be an ongoing source of consternation in the near future. It creates a charge that will be the power for an important dialectic on the topic. And, when all is said and done, having both the Harvard and the Wash U data sets with enormous quantities of information in each will allow us to ask important additional questions about the fundamental nature of the functional connectivity signals in the brain, and how they are affected between task engagement and default activity.