Algorithms and Informatics for Analysis of Neural Coding

Weill Medical Coll Of Cornell Univ, New York NY

Investigators

Linked publications & trials

Abstract

DESCRIPTION (provided by applicant): The proposed project will establish a multifaceted resource and collaboration to support a broad range of users, including experimental and computational neuroscientists, neuroinformatics developers, and others with an interest in the central problem of neural coding. The project supports as well the Human Brain Project initiative by incorporating linked neuroscience, informatics, and merged neuroinformatics goals. The neuroscience goal is an understanding of neural coding, synthesized via application of a wide range of algorithms to a database of neural data from multiple cortical areas, protocols, and preparations. Subgoals quantify the information present in patterns of visual neural activity, determine which features of activity carry information and transmit it between neurons and cortical areas, test contrasting coding hypotheses, and relate information and activity to behavior. Informatics goals include implementation and application of new analytic algorithms and the development, interfacing, and availability of a parallel computational resource, enhancing the value of a linked neurodatabase by adding value to archived data and encouraging submission and data sharing. The neuroinformatics goal tests different analytic algorithms against particular neurobiological processes, sites, and paradigms, including parallel channels formed by paired neurons. Neural coding is imperfectly understood. A main limitation is that at present, analyses and models are often derived from or tested on restricted numbers of datasets. To remove these limits, the project will link four disciplines: experimental neuroscience, analytical computational neuroscience, computer science, and informatics, via three aims. The first aim will develop, implement, and refine an array of algorithms for neural data analysis that probe how information is represented and processed. By applying several approaches to a given dataset, or one approach to many datasets, investigators can reach conclusions that are robust and not method-dependent. Lack of available computing power often limits the use of such algorithms. The second aim applies an existing 26-processor parallel computational array and designs neuroinformatics tools to ease data exchange with neurodatabases. Parallelized algorithms will thus be applied to a broad range of neural data. Utilizing insights into neural coding and parallel processing, the third aim will refine and advance algorithmic development, parallelization schemes, and neuroinformatic classification and exchange. New experimental/analytic collaborations will advance the design of protocols for the study of visual and somatic sensation that produces neural data--especially multiunit data--readily analyzed using evolving algorithms. Generated as well will be new ideas for hypothesis-testing analytics, all advancing collaborative computational neuroinformatics.