What’s Not Wrong with the BAM Project and How To Fix It

neuron4Hater’s gonna hate, or so they say, and the bloggerati have been lining up to point out the deficiencies (both real and imagined) in the mega-neuroscience Brain Activity Map project (the genesis of the plan is described by the New York Times here, Science here, and a Nature blog here). Many of these criticisms are thoughtful and reasonable, at least in response to the vague suggestions of the program so far outlined. But considering how little we actually know about the specific aims and funding, perhaps it’s worthwhile to take a moment to consider what is right about the proposal.

A Goldilocks Sense of Scale

As proposed in the original Neuron paper which laid out the plan (and a more recent sketch), and several appearances in the media by key figures in the plan, the aim is to study neural systems at the level of circuits (MIT Tech Review on the aims, CNN’s take here). The argument is that, while neuroscience has produced a great deal of insight from single-cell recordings that tell us about individual neurons, and another significant body of knowledge about the mass action of millions of cells as monitored by EEG or functional imaging, there is a significant gap in our knowledge of the important stuff in between. In contrast to the micro- and macro-scales where our efforts have been focused in the past, the BAM proposal looks to meso-scale analysis to provide crucial insight into the neural substrate of perception and behavior. At least one critic (who also happens to be a pioneer in the field) seems to consider talk of emergent properties as facile justification, but even without invoking that god of physics-envy we can point to hundreds of multi-cellular studies that have, with the small numbers they could record, shown that these techniques can answer very specific questions in many important systems. Just a few examples:  How do groups of cells represent planned movement? How do EEG recordings relate to neural firing patterns? How do rhythms coalesce from populations of cells, and how are those rhythms modulated? How does synchronous neuronal activity relate to perception or movement or epilepsy? How do neurons represent an animal’s physical space? How are patterns of synchrony related to plasticity and memory formation? How does neural activity driven by sensory input relate to ongoing neural activity? How do neurons multiplex different streams of information? How does circuit structure determine and constrain neural activity? Not only are these questions essentially opaque without meso-scale analysis, research in these areas provides an essential fulcrum for insight at larger and smaller scales. Further, for those concerned with the evolutionary and behavior validity of neuroscience (ethology) these techniques are even more important as scientists move toward experiments where animals are allowed to behave freely, rather than being immobilized/anaesthetized while being force fed the same atomistic stimulus over and over again.

The Timing

There will always be someone to argue that we don’t understand enough about the fundamental components of a system to study the global behavior of that system. Of course, the fact that such an argument can always be made doesn’t automatically make it right (as in the case of alchemy) or wrong (as in the case of genomics), so we have to look for evidence that proposed endeavor is poised for significant growth that could be catalyzed by coordination and funding. As mentioned in the previous posts, analytical techniques are being developed that can characterize activity from large networks using existing computational resources. Further, adding knowledge of the underlying connectivity can actually make these methods more tractable (and more useful). In terms of experimental methods, the possibility of combining the molecular ticker tape with individual cell barcoding has the potential to reveal connectivity and activity maps simultaneously. Yes, right now, those are untested techniques, which is why the BAM proponents reasonably suggest scaling up current technologies from less complex model organisms while developing the new tools. Critics are right to point out that understanding the roundworm (C. elegans) is not same thing as understanding the human. They are also right to point out that a disembodied brain activity map from a single individual, no matter how detailed, is useless without a deep understanding of the animal’s internal and external environment. Nonetheless, we should reject the argument that our inability to know everything about a system condemns us to never knowing anything about it (as pointed out also in this post on the Empirical Planet blog).

It is about Connectomics. It is about Behavior

The fact that the tagline about ‘recording all the spikes from all the neurons in a circuit’ has been used to sell the project (if that’s the catchiest they could come up with, the project is doomed), has led some commenters to suggest that BAM is only about recording a bunch of spikes from a bunch of neurons in some disembodied and unspecified brain mush. The outline of the program is still somewhat incomplete, but there is ample evidence that the planners have a reasonably contemporary understanding of systems neuroscience, which at least aspires to understand neural activity in the context of anatomical structure (note the Allen Brain Institute participation) and ethologically relevant behavior. Note that the Neuron paper proposes starting the project with C elegans. It’s no coincidence that this is the only organism for which the connectome is fully defined. In fact the authors specifically mention the tie to connectomics in their proposal, and one of lead authors (Rafael Yuste) has certainly taken anatomical connectivity very seriously in his work mapping circuits in the neocortex. In addition, if the BAM planners were unconcerned with behavior they wouldn’t have needed to suggest new techniques specifically capable of measuring neural activity in awake behaving animals. It could be argued that the Neuron paper implies a lack of emphasis on how brains become brains, or that it makes only a superficial connection to human disease models, but the authors are correct to imply that the potential benefit of circuit-level understanding of the (genetically accessible) mouse could produce tremendous advances in understand of both neurodevelopment and human disease.

The Aims

Many critics have argued that the aims of the project are too vague. For example, Leonid Kruglyak, has been a vocal critic on Twitter (@leonidkruglya), and was quoted in a piece on io9 suggesting that comparisons with the Human Genome Project (and it’s supposed 141x return on investment) is specious, because the HGP involved a comparatively well-defined coordination of sequencing research that was already going on (Story Landis from NINDS argues it’s actually easier in some ways). Others, like Donald G. Stein, as summarized in the New York Times, claim that the whole approach is futile (though the article doesn’t really give him chance to argue the case). The slipperiness of the project’s goals has also come under fire from Partha Mitra in a post on Scientific American’s site (and in blog comments here and elsewhere). He argues that the goal of ‘recording every spike from every neuron in a circuit’ is so fuzzy-headed, relying as it does on the ill-defined concept of emergent properties, that it is libel to lead us into a hopeless scientific bog (his clever metaphor, not mine). Mitra’s bonafides are impeccable, having spent a good long time doing the kind of systems neuroscience that is at the core of the BAM proposal (and written a book about it), but then stepping back to focus his attentions on underlying neural connectivity (the Brain Architecture Project).

Much of this criticism, however, is really based on the language the BAM proponents have used to sell their project, rather than the project as it would likely be implemented. No doubt this reflects a lack of marketing savvy for which no one else can really be blamed, but in actuality the project is as unlikely to fulfill its critics’ worst nightmares as its advocates’ best dreams. As for the lack of a clear hypothesis, I imagine the BAM authors would have to concede that point. They do, however, have a reasonably focused list of research questions like, “What is the functional connectivity diagram of a circuit?” and “How do local computations and long-range interactions influence each other?” It seems to me that if you believe in large-scale basic science at all (and I know that many people don’t), then you have to accept goals that are more broad that your typical NIH grant. Was the Large Hadron Collider driven purely by the hypothesis of the existence of the Higgs boson? (If so, then they can shut it down.) What was the specific hypothesis for the Human Genome Project? The moon mission?

The Smoke-filled Room

In terms of political optics, the initial planning for the BAM project has been a shit show. Many people see the project as essentially a Soviet-style boondoggle: a small cadre of connected individuals meet at an exclusive conference and hatch a multi-billion dollar plan to funnel money to like-minded scientists/friends, approved and integrated into the president’s upcoming budget proposal not because of scientific merit, but because of special access to White House science advisers. Some of the planners are not even working scientists (GASP)! While a portion of this criticism is over-the-top, it reflects a reasonable expectation of transparency and democratic process in science funding. And the political timing couldn’t be worse. When many NIH-funded scientists are facing sequester-driven budget cuts, the prospect of syphoning off a few billion for someone’s pet project (as it’s perceived) has not been well-received (Yuste responds to the zero sum argument here).

What To Do about It

Even if it true, as I suggest, that the administration’s missteps on this plan have been mostly political, if they hope for buy-in from the broader neuroscience community, then they will have to improve not only the branding but the actual process of planning and funding the BAM proposal. Here are some suggestions (in no particular order):

1)      The process needs to be transparent and open to broad participation. I don’t think I need to repeat the principles of open government here, but the administration clearly needs to be reminded. Centralized, top-down planning is absolutely not the approach people are hoping for. The funding mechanisms should allow for diverse projects with some entrepreneurial competition.

2)      The funding should not be scavenged from other projects. As many have pointed out lately, the prospects for early career scientists are increasingly grim, with NIH funding ill-matched to the numbers of recent PhD recipients, so drawing more money from that pool for a relatively specialized project…Let’s just say, “unlikely to engender broad support.”

3)      The tools and data must be freely accessible. Funding and participation from the private sector will surely require some allowance for ownership of certain discoveries, but we should encourage the kind of openness that has developed (at least in some areas of) in bioinformatics, which has simultaneously left ample room for profit.

4)      Don’t oversell it. The more talk there is of curing Parkinson’s or mapping the activity of human brains, the more smart people will call ‘bullshit.’ I’m not a science policy wonk, or a marketer, so I’m not really sure how to sell understanding the neural code to the general public, but let’s start by not alienating scientists (who can be a cranky lot).

 A Quick Review of My Previous Posts about BAM

My first post was an initial summary of the Neuron paper on which the proposal seems to be based. I followed with some of my thoughts on the appropriateness of the aims and practicality of the approach. In an effort to understand some of the proposed methods in more detail, I talked to a scientist who was a post-doc in Rafael’s Yuste’s lab and helped develop some of the high throughput techniques suggested for use in the proposal. Discussions with another scientist who has been working on new molecular recording techniques suggested profound possibilities for recording from and analyzing data from massive numbers of cells. Finally, I provide a quick summary of the more recent BAM outline that appeared behind a paywall at Science (though now linked above via Rafael Yuste’s website).

Photo Credit:  By GerryShaw (Own work), via Wikimedia Commons

~ by nucamb on March 10, 2013.

3 Responses to “What’s Not Wrong with the BAM Project and How To Fix It”

  1. [...] was said to be comparable to the HGP, which was estimated to have cost near $3B over ten years, so many critics were deeply skeptical about the possibility of drumming up $300M a year for systems neuroscience without sucking the life [...]

  2. [...] arguments of specific commenters about the BRAIN Initiative. I’ve certainly addressed some of the issues implicitly, and in fact I have agreed with some of the more serious criticisms of the project (such as [...]

  3. […] also been criticism that they are not doing enough to include women scientists in the process. As I’ve mentioned before, I still have my qualms about selling the project to the public based on promises to alleviate […]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

 
Follow

Get every new post delivered to your Inbox.

Join 48 other followers

%d bloggers like this: