Lots of people are writing about the Brain Activity Map, the proposed $3 billion Neuroscience project. This is my first post — color me skeptical. Here I’ll discuss one reservation.
As far as I an tell, the project is not hypothesis driven.
The project is pure technology: use big science to record every cell in the human brain. Start with small brains and, as the project moves along, do bigger and bigger brains, finally recording from every cell in the human brain. Then what?
I love Neuroscience and I love advances in technology, but to put so many bucks into one direction of technology, without a strong guiding hypothesis seems foolish. I have no doubt that as the project moves along many things will be learned. But without a guiding hypothesis, will it be worth the costs?
This evening’s NY Times has its third article on BAM. This is the first to express reservations. The article also describes the roots of the project in a meeting held in London in Sept, 2011:
For two days the scientists mostly “talked at each other,” he (Ralph Greenspan) recalled. Then George M. Church, a Harvard molecular geneticist who helped start the original Human Genome Project in 1984, said, “All right I’ve heard all of you say what you can do, but I haven’t heard anyone say what you really want to do.”
“I want to be able to record from every neuron in the brain at the same time,” Dr. Yuste replied.
Yes, Dr. Yuste, I want to do that, too. But why? What are the questions? I don’t hear a hypothesis. Without one, its impossible to do cost/benefit analysis. Cross your fingers, click your heels, will magic happen? A billion here, a billion there, eventually you’re talking real money.
The inability to create a central hypothesis is due to the primitive state of current neuroscience. We really don’t know how the brain works, how you go from the disparate firing of millions of neurons, to the cohesive nature of thought and action. My guess is that such a unifying hypothesis is possible, but that generating the hypotheses will not require recording all cells at once. Better lots of smart minds on smaller projects to build a hypothetical framework. Perhaps then a big, unifying project can and should be attempted.
I may be wrong. There may be a crucial set of important hypotheses to be tested. I’d like to hear them.
An interesting discussion, but I disagree. First of all I don’t think that a project like this needs to be hypothesis driven in the same way a NIH grant is. It’s a purposely visionary project that is meant to begin to answer many questions that are central to systems neuroscience. I could list a dozen of those specific questions and probably generate a good consensus on a subset of those from the the scientists involved in the BAM project (there’s a list in the Neuron paper). The Human Genome Project, for example was not hypothesis driven in the strict sense, but provided a framework for a huge body of research.
I’ve been blogging about BAM myself. First, with a synopsis of the Neuron paper (http://nucambiguous.wordpress.com/2013/02/19/neuroscience-emeril-style/), then a more detailed discussion of my thoughts on the project (http://nucambiguous.wordpress.com/2013/02/19/bam-my-thoughts-on-big-bucks-for-big-brain-science/) and finally with a synopsis of an interview with a researcher who helped establish some of the central techniques for the proposal (http://nucambiguous.wordpress.com/2013/02/24/bam-matters/).
nucamb, I’m of two minds. While I’d like to see the BAM project funded and proceed, I’m shocked that the magnitude, and frightened that it will crowd out other funding. I work close to the BAM area, and know the projects you’ve cited on your blog well. Its not clear that keeping funding constant, or increasing by a modest factor won’t be sufficient, or possible saturate the realistic need. I’m not convinced that smaller NIH projects should meet a criterion of hypothesis testing, while this much larger project doesn’t need to meet that challenge. Why? Finally, there is the mechanism. Why this project, not others? Why didn’t the govt say “we’re interested in a big Neuroscience project. Generate some proposals, and we’ll see if there are good ones”? Peer review is not great, but its better than anything else. BAM seems like a case of special pleading and connections, not judgement by a panel of peers.
Yes. I have the same questions. I’m torn between my basic agreement with the goals of the project and some of those same worries. If the guys in the smoke-filled room happen to come up with my candidate, should I withdraw my support? So, no it would not be worth it to fund BAM if it meant draining resources from a lot of other researchers. We still don’t know how the funding will be engineered (and with how much private support), so I’m willing to continue to say that I agree with the aspirational aims of the proposal.
I agree with John. No one can say if recording whole brain will answer all questions about its function. There is too many unknowns unknowns (to borrow words from D. Rumsfeld) so funding such project does not make any sense and it is waste of money.
I respectfully disagree. Recording every neurons is like sequencing every gene. There is no need for hypothesis because it’s obvious we’ll have to do it sooner or later.
Hello Prof. Kubie.
Just want to offer you my opinion on this and see what you think….
The scientific method cannot take sole credit for driving mankind forward. The empirical method has long played a crucial role in advancing scientific thought. Some of the most intricate projects ever designed were using empirical methods. As you know, scientific methods require a central hypothesis to drive experimentation, while empirical methods do not. Empirical methods depend on scaling (Froude/Reynolds numbers if it is a fluid), boundary conditions, and reproducibility. Continually running experiments on scaled models or observing the same phenomenon on identically functioning systems allows the experimenter to recognize patterns. While the scientific method relies on hypothesis-testing, the empirical method relies on observations and pattern recognitions (which I think in many ways the BAM project follows). Now you might wonder, why choose aimless observations and pattern recognitions over hypothesis testing (asking questions)? My answer is that I do not believe hypothesizing (or asking questions) is a fundamental and potent behavior we possess compared to pattern recognition. Asking questions is a higher order function (requires higher ordered circuitry), and therefore prone to great differences and faults within the systems (the higher the order of the circuit, the greater the std-dev and tolerance) . For us to formulate the right question about a certain experiment means that the experiment must be within our scope of thinking – this is not necessarily the case, this is why we find quantum mechanics counter-intuitive. What we are good at is, pattern recognition – this seems to be a fundamentally hardwire in our brains. We may not be able to formulate questions well, but we are very good at recognizing patterns! This is why we have had such great success in engineering and fluid mechanics, at least (I can give you many examples at a later time).
So why the BAM Project? It is mostly likely going to be plagued by cost overruns. In this case, I ask you to consider the Carnot cycle efficiency peaking at roughly 30%…I think the BAM Project operates within this efficiency, and I have observed many projects/programs/agendas operate at least one order of magnitude less than this thermodynamic standard (you can break a lot of laws in life, but you can’t break the laws of thermodynamics!). As for the benefit, this project needs to be done eventually, without a central hypothesis. Why? It’s been observed that neural networks function properly using noise (chaos) as a tool for filtering and long-term potentiation. Noise is something we cannot hypothesize on, it is purely dependent on the inherent stochastics of the system and the only way to understand it is by pattern recognitions. There is this great book called Theoretical Neuroscience by Dayan and Abbott, which shows the necessity of noise on neural systems to maintain filtering rate/potential (a key component in neural function). I will end here….See you in class!