Neuroscience in a Global Era: What Are the Best Ways to Collaborate?
The past five years have seen an upswing in large-scale efforts that aim to solve the brain’s most challenging enigmas. The Obama administration’s BRAIN Initiative and the European Commission’s Human Brain Project, both launched in 2013, represent massive and sometimes controversial investments in neuroscience that account for a huge portion of funding in the field. Is it money well spent? Or has large-scale neuroscience been oversold?
This month, the journal Neuron explores the promise and pitfalls of large-scale neuroscience in a special issue on global neuroscience. In a series of short comments, a cadre of neuroscientists, including a handful of investigators involved in the Simons Collaboration on the Global Brain, weigh in on when large-scale collaboration is most useful, how to encourage data-sharing, and the need to maintain individual creativity. A related commentary in Nature, by SCGB investigators Zachary Mainen and Alexandre Pouget, along with Michael Häusser, outlines a plan for so-called mesoscale collaborations with a bottom-up, rather than top-down, organization.
One theme that emerged from the series is that in the era of big science, small or medium-sized collaborations are still important. Adrienne Fairhall, a neuroscientist at the University of Washington and member of SCGB’s executive committee, argues that smaller groups are more flexible and creative, properties that are particularly important in early-stage projects:
“I don’t immediately expect large group efforts to replace the vital role of small-team creativity. The complexities of hammering out an initial idea and working through its experimental validation may be better done in the flexible environment of a smaller group. Where I see a great opportunity for large-scale teamwork is in the development, enhancement, validation, and deployment of new technologies and the streamlining and standardization of procedures.”
Richard Tsien a neuroscientist at New York University, advocates for ‘medium science,’ where networks of labs work together like a small village:
“I like ‘medium science,’ performed by multiple labs each doing small science but working in loose federation. It requires heaps of communication — for example, a weekly group meeting where unfinished work is presented to the whole institute. Trainees get extra counsel from their advisor’s close colleagues, acting like aunts and uncles in a pride of lions. Camaraderie, self-interest, and frequent visitors help spread new methods. … For most of neuroscience, medium-sized structure still delivers the best combination of focused training and diversity to spark innovation. If big science is needed, why not incorporate family-like small science and village-like medium science as components? Even large international programs can benefit — human nature generalizes across national cultures and scientific arenas.”
Data-sharing has been one of the biggest hurdles to collaboration. SCGB investigators Karel Svoboda, at the Howard Hughes Medical Institute’s Janelia Research Campus, and Matteo Carandini, at University College London, highlight the need for standardization across labs to enhance sharing. “Concept-driven collaborations might further arise naturally with standardization of experiments and data formats (e.g., NWB),” Svoboda writes. “If several laboratories working on related problems share data for collective mining, a deeper understanding will emerge.”
Carandini notes that the study of behavior, in particular, would benefit from standardization:
“Laboratories that study behavior do so using different tasks, recording from different neurons, in different brain regions, in different animals, in different species. Such disparate observations can neither be pooled nor compared. One solution is to get multiple laboratories to focus on a set of standardized tasks — in highly controlled conditions such as virtual reality — and focus on the mouse, with its arsenal of genetic tools and databases. Such standardization would make it fruitful to pool and share the resulting data.”
David Anderson, an SCGB investigator from the California Institute of Technology (Caltech), concurs. He argues that the field needs to develop ways to measure behavior that have a spatial and temporal resolution similar to that for measuring neural circuit activity. Advances in machine learning can help standardize tracking of behavior. Without that, “it will be difficult to know whether the codes, algorithms, and representations that neuroscientists can extract mathematically from patterns of neuronal activity are actually extracted by the brain. Advances in machine vision and machine learning are creating exciting opportunities, at the interface of computer science and biology, to develop standardized, robust, and broadly applicable methods for measuring behavior and integrating those measurements into platforms for large-scale recording of neuronal activity.”
Mainen, Pouget and Hausser’s Nature commentary echoes many of these sentiments. They argue that focused grass-roots collaborations will provide the best antidote to the technical and sociological barriers of ‘big’ neuroscience. These collaborations would start small and grow from the ground up, driven by investigators’ own shared interests rather than external directives. The authors outline several specific steps that they say will be essential for the success of these collaborations.
Focus on a single brain function
“The common goal has to be ambitious, yet reachable within, say, ten years, and well defined. A whole-brain theory of one brain function — a single behaviour — could meet those requirements. If a collaboration were largely limited to labs interested in the same behaviour — such as courtship in fruit flies, or foraging in mice — clear, shared objectives could be defined at the start.”
Combine experimentalists and theorists
Neuroscience labs often specialize in either experimental or theoretical approaches. Although the two groups “often meet at conferences to share ideas, they rarely converge when it comes to the design and interpretation of experiments. So, concrete steps are required to catalyze more meaningful interactions, such as embedding theory Ph.D. students and postdocs in experimental labs and vice versa.”
Assign credit in new ways
A major hurdle in standardizing data and encouraging collaboration stems from the competitive and individualistic nature of neuroscience. The authors suggest taking inspiration from genomics, which since the Human Genome Project has developed a much more collaborative structure, and particle physics, which has long been centered around huge collaborations. The ATLAS collaboration at CERN, a particle-physics lab in Switzerland, provides a successful example of “a distributed model, avoiding a central command-and-control structure.” The authors say it “characterizes one of the largest and most effective mega-projects.”
Collaboration within the SCGB
The SCGB is already making headway in one of these areas, the collaboration between theorists and experimentalists. Indeed, the collaboration was built around this idea. “The SCGB is a great catalyst for bringing together mathematicians and neuroscientists in a serious way, not just to hear each other talk,” says Markus Meister, a neuroscientist at Caltech and SCGB investigator.
For example, at Caltech, Meister, David Anderson and Pietro Perona are studying how animals respond to threats. A mouse that sees a looming circle above him will think he’s about to be attacked by bird and either freeze or run and hide. How does the animal make that choice? Each investigator is exploring the problem from a different angle. Meister is analyzing how the mouse uses visual information to determine whether it’s under threat. Anderson is studying how the animal’s internal state — such as anxiety or fear — influences its decision. Perona is developing mathematical models that show how animals incorporate prior experience and ongoing sensory input to make decisions. The researchers meet weekly to discuss their progress and ultimately hope to create a cohesive picture of how an animal’s emotional state and incoming sensory information come together to direct its behavior. “Having three different viewpoints has been inspirational,” Meister says. “It’s enlightening in how you choose to proceed.”
