Akul et al discuss this in a recent editorial, and the main challenges they touch on are: 1) how can we link tiers of analysis such as macro- and micro-connectomes?; 2) can we uncover how distinct gene mutations lead to a similar clinical phenotype by integrating info?; and 3) most generally, what are the best tools and methods to combine our disparate data types: genetic, cellular, anatomical, electrophysiological, behavioral, evolutionary, and computational?
The Neuroscience Informatics Framework is the place to go to find links to sources of raw data. For example, there are currently 145 public microarray data sets available, most of which compare expression between different conditions or different brain regions. The search is OK, but, as the authors urge, it could become better if up-loaders recognize that the most important thing they can do is make their data machine-readable. Although not explicitly mentioned, what I noticed most in the article was a tremendous amount of excitement.
Huda Akil, et al. Challenges and Opportunities in Mining Neuroscience Data. Science 331, 708 (2011); DOI: 10.1126/science.1199305