By Joe Alper

April 15, 2003 | Neuroscientists are calling for multidimensional public databases to better coordinate advanced brain research, whose fragmented efforts are currently drowning in data.  Without such databases, neuroscience will continue to lag behind other areas of biomedical research that have benefited from advances in bioinformatics, thwarting attempts to make a quantum leap in understanding all aspects of brain development, function, and disease.

“If we are ever going to come to a thorough understanding of how the brain functions and how to treat the diseases of the brain, neuroinformatics will have to play a central role in those efforts,” says Stephen Koslow, who heads the National Institute of Mental Health’s Human Brain Project. To help stimulate interest in neuroinformatics, Koslow organized a two-day session titled “Neuroinformatics: Genes to Behavior” at the American Association for the Advancement of Science (AAAS) meeting held in Denver in February.

Since the day when then-President George Bush declared the 1990s to be the “Decade of the Brain,” neuroscientists have made significant strides in understanding how the brain is organized structurally and about the inner workings of its individual components, the nerve cells that convey messages throughout the brain and central nervous system. But efforts to integrate this information into a coherent picture of how the brain generates thoughts and actions have hit a brick wall.

The lack of good informatics tools to catalog and analyze neuroscience data is “paralyzing the entire field of neuroscience,” said Floyd Bloom, chair of the department of neuropharmacology at The Scripps Research Institute and CEO of Neurome Inc., a neuroinformatics startup.

What’s needed, said Bloom and other speakers, is a concerted effort to develop multidimensional databases that can handle the enormous dynamic range of data generated by neuroscience research. At one end of the range are the molecules that transmit electrical signals throughout the brain. Next, in terms of scale, come single neurons, followed by neural networks, and, finally, activity of whole brain regions.

“Integrating data across these different domains is where the real challenge lies, and it’s where bioinformatics comes in,’ Bloom said. “We need new technologies to harvest information from all of our data, just as the Human Genome Project required new bioinformatics tools to accomplish its goal.”

What neuroscientists would like is a public database along the lines of GenBank, where researchers would be able to both deposit and extract data in a standardized format. Toward that end, the Human Brain Project has funded a number of 10-year efforts to develop model databases.

One of those projects, SenseLab, involves developing a comprehensive approach to building integrated, multidimensional models of neurons and neural systems using the olfactory pathway as the focus. Gordon Shepherd, who heads SenseLab and is a professor of neurobiology at Yale School of Medicine, told the AAAS symposium that GenBank and “BrainBank” have one major difference:

“While GenBank deals with linear sequences of genomic data, a neuroscience data bank would include complex structures of cells and their organelles; physiological recordings of complex activity occurring over different time scales; and brain scans from different human and animal subjects. This means that we have to develop bioinformatics tools that will allow us to put the data through complex procedures for alignment in order to be able to make useful comparisons among those data.”

Shepherd and an international group of colleagues have been developing such tools and have used them to create two sets of three interconnected databases to handle different types of data. One of the three olfactory databases contains information on the more than 4,000 olfactory receptor genes and proteins, including the recently identified complete olfactory genomes of the mouse and human.

A second database holds data on the odor molecules that interact with these receptors. The third houses odor maps -- images of the neural activity that occurs in the brain when a particular odor molecule stimulates a given receptor. Each database is linked with the other two, providing, according to Shepherd, “a way to study how odor space is represented as neural images in the brain.”

The second suite of three databases supports experimental and computational analyses of membrane properties, such as receptors, channels, and neurotransmitters, of neurons involved in processing odoriferous stimuli. A cell-properties database supports a proteomics approach to the membrane properties found in different types of neurons in the olfactory system.

Data about a newly identified gene or protein in a given cell type in a given region would be deposited here. At a more detailed level, the neuron database represents those membrane properties as occur in different combinations in different parts of a particular neuron. This database also provides search tools that enable an investigator to search across neurons for various combinations of properties.

“This is the first step toward enabling neuroscientists to identify complex motifs of properties that are shared by different neurons, which is central to identifying what role they play in the nervous system,” Shepherd said. The final database allows researchers to integrate the data from any or all of the other databases into any one of a collection of more than 80 computation models of different types of neurons and their properties with uniform formatting and curating.

Other speakers at the AAAS symposium outlined the beginnings of their efforts to create databases of other types of neuroscience data. For example, Arthur Toga, head of the UCLA neuroimaging laboratory, spoke about his group’s efforts to develop a human brain atlas, while Michael Gazzaniga, director of the Center for Cognitive Neuroscience at Dartmouth, spoke about his work cataloging data sets from functional magnetic resonance imaging studies of how the brain works while thinking.

All of the speakers bemoaned the fact that they have yet to get computer scientists to buy into neuroinformatics in the same way that they’ve managed for the larger bioinformatics effort. But they also noted reluctance among the neuroscience community to become involved in what must become a communitywide effort. Shepherd put the latter problem this way: “Nearly everyone who has tested SenseLab has been impressed, but none of them are lining up to add data to it.”