Gast ::
| Hauptseite > Publikationsdatenbank > Semantically-enriched description of electrophysiology data analysis workflows |
| Hauptseite > Publikationsdatenbank > Semantically-enriched description of electrophysiology data analysis workflows |
| Poster (After Call) | FZJ-2024-04031 |
; ;
2024
Abstract: Extracellular electrophysiology is a common experimental technique to investigate brain function. Also, the outcome of brain simulations on the neural network level can be related to electrophysiological measures obtained from such experiments. The analysis of electrophysiology data requires specific methods of varying complexity, which are frequently implemented in computational workflows that take the form of a series of scripts that read input datasets and produce result files [1]. There are challenges to describe the processes throughout the workflow: (i) parameters are often selected iteratively, and the subsequent results depend on the details of the iterations; (ii) finding result files in a collection is difficult, as several parameters may be stored in a non machine-readable format inside files or using non-standardized names; (iii) there are several variations of analysis methods that can be applied to the data with similar purposes (e.g., different algorithms to compute the power spectral density from local field potentials); (iv) an analysis method can be implemented by different software codes (e.g., toolboxes such as Elephant [2] or MNE [3]; see also [4]) that adopt different names for the functions and their parameters. In the end, this produces a scenario where it is difficult to find, understand, and compare analysis results, especially in collaborative environments where large results sets are available in shared repositories. To overcome those challenges, we developed a framework to generate machine-readable descriptions of the workflow execution that are enriched with the relevant semantic information. The details of the inputs, outputs, and parameters of the functions called within the workflow scripts are captured with minimal user intervention using the Alpaca (Automatic Lightweight Provenance Capture; RRID:SCR_023739) [5] toolbox. This produces a detailed record of the atomic steps used to generate an analysis result. The provenance information is enriched with annotations using the Neuroelectrophysiology Analysis Ontology (NEAO), which we developed as a unified vocabulary to standardize the descriptions of the methods involved in the analysis of extracellular electrophysiology data. We show real-world examples where the framework was used to generate machine-actionable descriptions of different analyses of an electrophysiology dataset and highlight how it is possible to query information, facilitating finding and obtaining insights on the results (e.g., using knowledge graphs). In the end, this approach improves the analysis workflow by making the details of the results known and standardizing their description. Ultimately, we discuss how this methodology can be applied more generally to computational workflows (e.g., in network simulation or AI applications) to help representing the results according to the FAIR principles [6]. REFERENCES [1] M. Denker and S. Grün, LNCS, 10087, 58, 2016. [2] M. Denker et al., Neuroinformatics 2018, P19, 2018. [3] A. Gramfort et al., Front. Neurosci., 7, 267, 2013. [4] V. A. Unakafova and A. Gail, Front. Neuroinform., 13, 57, 2019. [5] C. A. Köhler et al., arXiv, 2311.09672, 2023. [6] M. D. Wilkinson et al., Sci. Data, 3, 160018, 2016.
|
The record appears in these collections: |