Organizing Metadata of Complex Neurophysiological Experiments

Technological progress in neuroscience allows recording from tens to hundreds of neurons simultaneously, both in vitro and in vivo, using various recording techniques (e.g., multi-electrode recordings) and stimulation methods (e.g., optogenetics). In addition, recordings can be performed in parallel from multiple brain areas, under more or less natural conditions in (almost) freely behaving animals. Consequently, electrophysiological experiments become increasingly complex. Moreover, to disentangle the relationship to behavior, it is necessary to document animal training, experimental procedures, and details of the setup along with recorded neuronal and behavioral data. Given these various sources of complexity within an experiment, the availability of such information about the experiment, commonly referred to as metadata, is of extreme relevance for reproducible data analysis and correct interpretation of results. Typically, experimenters have developed their own personal procedure to document their experiment, allowing at best other members of the lab to share data and metadata. However, at the latest when it comes to data sharing across labs, details may be missed. In particular if collaborating groups have different scientific backgrounds, implicit knowledge is often not communicated. In order to perform interpretable analysis of the data, each data set should therefore clearly link to metadata annotations about experimental conditions such as the performed task, quality of the data, or relevant preprocessing (e.g., spike sorting).In order to provide metadata in an organized, easily accessible, but also machine-readable way, an XML based file format, odML (open metadata Markup Language), was proposed [1]. Here, we will demonstrate the usefulness of standardized metadata collections for handling the data and their analysis in the context of a complex behavioral (reach to grasp) experiment with neuronal recordings from a large number of electrodes (Utah array) delivering massively parallel spike and LFP data [2]. We illustrate the conceptual design of an odML metadata structure and provide a practical introduction on how to generate an odML file. In addition, we offer odML templates to facilitate the usage of odML across different laboratories and experimental contexts. We demonstrate hands-on the advantages of using odML to screen large numbers of data sets according to selection criteria (e.g., behavioral performance) relevant for subsequent analyses (see companion posters by Denker et al.). Well organized metadata management is a key component to guarantee reproducibility of experiments and to track provenance of performed analyses