Speaker
Description
A fast and reliable data processing is becoming increasingly crucial in biological and medical applications for extracting quantitative information and providing reliable predictions, especially when coming to AI-assisted tools. At the same time, several concerns may be raised regarding the analysis of human data: fair comparison between heterogeneous data sources, reliability and explainability of data processing outcomes, potential clinical decisions based on them, and an acceptable trade-off between reproducibility in the analysis and privacy guaranteed by National and European regulations.
In this framework, the experience of Cobrawap (Collaborative Brain Wave Analysis Pipeline, https://doi.org/10.5281/zenodo.10198748) for the standardized multi-modal analysis of brain activity, either experimentally recorded or simulated, has provided interesting solutions to these problems, complying with FAIR and Open-Science principles. Showcased as a successful interdisciplinary project since the Human Brain Project, then within EBRAINS and the PNRR initiative EBRAINS-Italy, Cobrawap builds on the co-design between the needs expressed by the neuroscientific community and the technological solutions offered by our community, namely bringing into the neuroscience domain the computational expertise developed by INFN during decades of massive data analysis for core-business experiments.
We will show the latest technical solutions implemented with the purpose of releasing Cobrawap as a ready-to-go service for users coming from different backgrounds, leveraging either on-premise (e.g., HPC execution on CINECA facilities) or cloud solutions (e.g., EBRAINS Collab), still offering a customizable experience thanks to the modularity of the pipeline, and its innate collaborative design. The backbone execution is left behind the curtains, orchestrated by workflow managers such as Snakemake or CWL, with the final users in charge of handling simple human-readable configuration files.
Future outlooks regard the use of Cobrawap for the study of artificial neural networks, with the two-fold aim of unveiling the “black box” and offering biologically inspired solutions to enhance the accuracy and the energetic performances of current state-of-the-art AI architectures.