NIX – Neuroscience information exchange format

I am a developer of SpikeInterface and worked with the Nix team to integrate their format into our framework. From my experience in electrophysiology and spike sorting, I firmly believe that standardised, universal file formats are important for reproducible analysis and multi-lab collaborations. Nix's general and flexible design makes them an excellent option for satisfying this need and I would be happy to see them endorsed by the INCF. I also agree with Andrew Davidson's comment that having competing standards, especially ones with large design differences, is valuable, therefore, I would like to see both Nix and NWB be integrated into the larger neuroscience community. READ LESS

I was a chair person of the Dynamic Brain Platform, INCF Japan Node, which has promoted studies on the dynamic principles of brain functions through unifying experimental and computational approaches in cellular, local circuit, global network and behavioral levels. My recent missions is to build a workflow with NIX/odML for human multi-dimensional data and an extended system based on NIX/odML associated with OWL/SWRL schemes as a semantic web technology for a large-scale neuroscientific data in the research project on "data-driven ontology discovery" in the framework of collaborative research project with RIKEN Center for Brain Science (RIKEN CBS). I expect further developments of NIX related frameworks continuously and I will contribute to a part of the effort. READ LESS

We needed solution of storing the neuroinformatics laboratory experiments of the University of West Bohemia. An analysis of available data standards for storing neurophysiological experiments is done  where each data standard is evaluated on the FAIR principles of storing open science data. The most appropriate standard (NIX) is selected based on the FAIR principles assessment and the complexity of potential conversion. Our student implemented a conversion tool to convert the current experiments to the new data standard. The G-Node’s GIN repository was selected as the most convenient way to store the newly selected NIX data standard, replacing the EEGbase portal. The version of the repository hosted at G-Node’s servers was chosen as the primary storage of the data, as the offered DOI service, which is paid otherwise, greatly supports the findability principle. READ LESS

We are building our data management and research on top of open source tools such as Neo and odML. In this combination, Nix plays an important role in our workflows as it is one of the few formats to store not only electrophysiological data, but also complementary data types, in an open, generic and non-commercial format. Most importantly, the Nix format is extremely flexible and therefore fits many use cases. Nix is supported as format by the Neo library, providing a representation of the data that is independent of the data source. This Neo IO was mostly implemented by the G-Node and is continuously updated based on the latest Neo development as well as user feedback. The design of Nix allows us to combine data and extensive metadata (describing intricate details of the data) in a single framework, permitting the consistent storage of both.Thus, no references between data and metadata files are required, making the distribution of datasets among collaborators more robust. As principle developers of the Elephant data analysis package, we currently see Nix as the primary format for exporting analysis results in a way that it includes corresponding analysis metadata (e.g., axis labels, analysis parameters, etc). Nix also forms a basis to apply analysis tools from nearby fields, as we hope its generic structure will allow for an easy import/export of the analysis data from different frameworks. READ LESS

A few years ago, my laboratory at Cornell University (Ithaca, NY, USA) initiated a set of software projects for which we required a data model for the analysis and archiving of various data and metadata.  Despite the multiple types of data to be acquired and interrelated, we preferred to work with a single data model across projects.  Ease of use and strong Python support were essential.  Most important, however, were flexibility and futureproofing:  the data model should be open and minimalist enough so that we could encode our unusual data structures efficiently and expect them to be easily readable into the foreseeable future without resorting to legacy or customized importers.  After some consideration we decided to use NIX for our projects.

NIX has since become our go-to data format for any large lab datasets.  The NIX API is minimalist and sufficient for all of our needs without being inordinately complex – it feels very natural and it is easy to decide how to store data within a file.  It is certainly preferable to vanilla HDF5, as it collapses the huge HDF5 design space in terms of how to lay out data without losing the power to store arbitrary data structures.  (In terms of future-proofing, I also note the prospect of NIX transparently supporting non-HDF5 back ends, such as a native filesystem, with potential benefits for the durable archiving of very large datasets). 

Our Ceed project (https://github.com/cplab/ceed) manages experiments in which complex, arbitrary spatiotemporal patterns of light are delivered to an optically sensitive brain slice from a transgenic mouse, while recording electrophysiological activity in the slice from a 120-channel planar multielectrode array.  The objective is to end up with a data file including the optical stimulation state, electrophysiological data, and other experimental parameters (e.g., perfusion switching, metadata) aligned correctly in time.  Ceed generates the user-defined spatiotemporal light patterns and records them as data in NIX while the electrophysiological data are acquired in a proprietary commercial format optimized for streaming.  After the experiment, we merge these electrophysiological data into the NIX file, which then is used for all analyses and archiving. 

Glitter (https://github.com/matham/glitter) is software we created for manually coding behavioral data from video files, or for editing and further annotating automatically-scored video data (e.g., XY tracking) so as to generate a single file containing all encodings based on a given video file.  All encoding data are saved into a NIX file which is subsequently used for analysis and archiving. 

In our neuromorphic models, we also use NIX to store realtime debugging data of model state and the explicit output of neuronal and network state variables. 

We also are planning to use NIX for cell counting, regional localization, and classification data derived from three-dimensional image files of optically cleared mouse brains registered to the Allen CCFv3 and mapped with respect to a few different candidate atlases.

The flexibility and minimalism of NIX is central to our needs, and is well suited for the creation of multimodal data files generated from complex experimental studies involving (for example) diverse physiological and behavioral data with experimental control.  It enables common data types to be written easily in standardized formats without hindering the encoding of new or specialized data structures.  It is just what we need in a data model for open neuroscience, and I recommend its endorsement by INCF as a standard. READ LESS

I am one of the principal developers of Neo, which provides a standardized in-memory representation of electrophysiology and optophysiology data in Python, together with support for reading a wide range of neurophysiology file formats. Until recently, Neo users had very few choices of high quality formats for writing neurophysiology data. As I commented in the community review of NWB:N, the Neo developers are very happy to see the emergence of both NWB and Nix, and we hope that both will be endorsed by the INCF. In this case I think it is valuable to have two somewhat competing standards, as the use cases for such formats can be highly complex, and it is very difficult to judge in advance which format will be optimal for which use case. Furthermore, NWB and Nix are at opposite ends of the design space, the former being highly opinionated about the embedding of the experiment structure and metadata into the file structure, the latter being highly generic and flexible. READ LESS