Collaborative HPC-enabled workflows on the HBP Collaboratory using the Elephant framework

The degree of complexity in analyzing massively parallel, heterogeneous data from electrophysiological experiments and network simulations has reached levels where novel tools forming workflows for managing data, metadata acquisition, pre-processing, and analysis in a reproducible fashion are in high demand. Moreover, this complexity calls for new conceptual approaches in organizing scientific work. The nature of these new, highly manifold projects requires work to be performed in larger, multi-disciplinary collaborations. The collaborators that need to interact closely require the supported from powerful computing resources, especially when dealing with large and diverse data sets.The Human Brain Project (HBP, http://humanbrainproject.eu/) aims at creating and operating a scientific research infrastructure for the neurosciences to address such needs for integrative software environments. At its core, the HBP features the Collaboratory (http://www.collab.humanbrainproject.eu), a web-based platform to jointly implement research projects. It enables the ability to create so-called “Collabs” that consist of a joint workspace and associated apps (see also poster by van Papen et al). An app provides the functionality to launch Jupyter notebooks, which is in particular interesting for interactive data analysis. Powerful as this approach appears on paper, it is unclear how these developments translate into implementing complete analysis workflows in a real-world collaborative analysis scenario. Here, we show how diverse tools, from software development, general science and software specific for neuroscience can be successfully combined in a to form a collaborative analysis workflow hosted on the HBP Collaboratory. Three emerging open-source software tools from neuroscience represent the tool basis which builds the analysis pipeline: (i) data of different origins are represented in a standard form using the Neo framework, (ii) complex metadata accumulated in relation to an electrophysiological experiment are managed using the open metadata markup language (odML) and (iii) analysis is performed utilizing the Electrophysiology Analysis Toolkit (Elephant, http://neuralensemble.org/elephant/). The Elephant tool is a recent community-centered analysis framework for the analysis of multi-scale high-dimensional activity data. Elephant is a modular software component that provides generic library functions to perform standard and advanced analysis. All these domain-specific tools are augmented by generic tools, such as version control systems or workflow management solutions, to form a blueprint for performing interdisciplinary, collaborative work including access to high-performance computing facilities for advanced, but computational demanding analyses