Neuroinformatics: Cleaning up the mess we’re all making

Despite what we like to tell ourselves, doing neuroscience is…not an exact science. We say that reproducibility of results is key. But the truth is that there’s a nauseating amount of variability across labs in every step of the scientific process. Data is collected with different brands of equipment, or tools made in-house. Different algorithms for cleaning and filtering results can turn even identical raw data into vastly different datasets after only one stage of processing. Assorted file formats, unique record keeping procedures, and a lack of commonly accepted nomenclature can make sharing data a technical nightmare. This makes the notion of any lab exactly replicating another’s results a near impossibility.

Enter Neuroinformatics, a sort of meta-field of neuroscience that tries to standardize the way we collect, label, and share the fruits of our labor. The boundary between neuroinformatics and computational neuroscience isn’t always clear, since according to the International Neuroinformatics Coordinating Facility, computational neural modeling falls under the banner of “integrating research findings”. Suffice it to say, neuroinformaticists are trying to battle the data deluge and disorganization that currently plagues our field. And I, for one, couldn’t be more grateful. In fact, I just spent some time hanging out with these fine people at the INCF 2012 conference in Munich. Here are some of the great tools I found out about there:

Data and Code Sharing

CARMEN  – With a clean and simple interface, this data-storing space from the UK is a platform for neurophysiologists to keep and share their data and analysis code.

CRCNS – This site collects and shares neurophysiological and eye movement data for the purpose of informing computational models. They are also working on a standardized methodology for describing and sharing stimuli, both visual and audio.

Information Databases

NIF – The motherload. This NIH-backed super-database has a search tool that allows simultaneous searching across an array of biochemical, genetics, and imaging databases, both for literature and primary data. Their NeuroLex initiative also helps determine synonyms for brain areas which can lead to more accurate search results. Its size makes it a little unwieldy to use, but it definitely provides a lot.

BrainInfo – Primarily a neuroantomy source, this easy-to-use site allows you to search a specific brain area and get a variety of information on it. They also provide mapping for primate and rodent data.

INV-BF – This Japanese initiative includes imaging, anatomical and systems data for a variety of invertebrates, which makes it a good resource for lesser-studied organisms.

NeuroElectro  – The goal of this newly-launched site is to compile the neurophysiological properties of a variety of neuron types into one central database.

Spikesorting Evaluation

G-node  – I am hugely supportive of the current push to share, standardize, and actually test spike-sorting algorithms. This project provides simulated raw data which can be run through your spike-sorter of choice and then compared to the ground truth spike data. It doesn’t actually provide different algorithms, but hopefully the successful ones will get published and compiled elsewhere.

Online Atlases/Visualizations

Virtual Fly Brain  – A wealth of imaging data from everyone’s favorite insect is collected and displayed in a very attractive format in this virtual atlas.

3-D Brain*  – This site offers a variety of brain atlases to choose from, which can then be viewed through their 3-D constructor

OpenWorm*  – This is just fun, I don’t care who you are. The knowledge of the C. elegan’s entire neural wiring diagram is utilized to make an interactive platform for exploring anything you could want to do with the little worm’s anatomy.

*These sites tend to not cooperate with Firefox. I suggest Chrome.

It may seem like neuroinformatics is mainly concerned with housekeeping details, and perhaps is only of passing interest to the rest of the neuroscience community. I would argue that this is a very misguided view. How we choose to store and share our data will determine the pace of progress in this field. New discoveries are all about connecting previously disconnected ideas. Ignoring the need for a proper infrastructure just adds extra obstacles to this path. Only with the right tools can we make our work as efficient and accurate as possible. And for that reason, I think we should all take up the banner of neuroinformatics by utilizing the means that are currently provided, and working to shape them for the future.