CARMEN: Distributed Search and Data-Mining in Complex Neurophysiology Data

Abstract

Event Abstract Back to Event CARMEN: Distributed Search and Data-Mining in Complex Neurophysiology Data Bojian Liang1*, Tom Jackson1, Martyn Fletcher1, Mark Jessop1 and Jim Austin1 1 University of York, United Kingdom We describe research within the UK CARMEN [1] e-Science project which is addressing the challenge of creating and managing experimental data and methods within the context of Neuroscience research. The traditional research approach of testing a hypothesis and publishing the results is hampered in situations where others need to build on the results and need access to the data or original methods. CARMEN addresses this issue by allowing scientists to share data & methods within a collaborative Grid environment. The CARMEN platform (a CAIRN) is a Grid-based, shared data and services repository. Central to the data management challenge is providing visualisation and search capabilities across the stored, high-volume, time-series, spike train data from nerve cell recordings. A prior UK e-Science project, DAME[2], developed a distributed signal search engine, called Signal Data Explorer (SDE) [3]. Interoperability between SDE and other neuroscience software services is being investigated within CARMEN, to provide a centralised capability for visualisation and search. The SDE provides data visualisation, transformations and real-time search capabilities for complex time-series signal data. The interactive search capability is provided by the SDE via distributed Grid data repositories which use: • Distributed data management - SDE interfaces directly to the Storage Resource Broker (SRB) from San Diego Supercomputer Center (SDSC); • The pattern matching architecture consisting of Pattern Match Controller (PMC) and Pattern Match Engine (PME) components. The raw data to be stored by the CARMEN CAIRN nodes consists of spatiotemporal signals expressed, either as time-series recordings (collected by single electrodes or MEAs), or as image files. Querying of the data at the “signal” level is provided via a highly intuitive and interactive SDE GUI. This GUI allows users to formulate queries by simple selection of regions of interest in time-series signals, or via libraries of templates queries, built up over the course of prior experiments. The SDE GUI also supports complex visualisation requirements, permitting users to explore data events and features in detail or in a broad overview, as required. Auxiliary viewing windows permit users to arrange collated views on data signals, and support additional functionality such as signal superimposition, filtering and spike sorting. The talk will provide a demonstration of progress on the development of the search and visualisation capabilities, working on experimental MEA data, and will show how additional analysis functionality, such as spike sorting or correlation functions, can be supported within the tool. User feedback is an important aspect of the development iteration and we will conclude with examples of how SDE is supporting laboratory experiments within the CARMEN consortium.