Challenge 2: Brain-on-Chip NeuroCAAS
In the Brain-on-Chip project, neuronal circuits are reconstructed in vitro from human induced pluripotent stem cells (hiPSC’s) to facilitate biomedical experimentation purposes such as diagnosis, stimulation and treatment effect analysis. These cultures are plated and matured on high density CMOS multi-electrode arrays (HD-MEA’s) – another type of MPS – after which their electrogenic activity can be recorded. Spike sorting is the process in which these voltage traces – which are spatially and longitudinally correlated – are converted to spiking activity of single neurons. Many spike sorting algorithms exists, and contain various parameters that can be tuned to manipulate their outcome. Nevertheless, the raw outcome of the spike sorting algorithms has to be manually curated before analyses can happen. The subsequent analyses allow for examining single cell and network behaviour using a number of analytical tools.
A NeuroCAAS is a cloud analysis platform that moves the arduous process of data processing, sorting, analysis and interpretation to a specialized environment that can run several sorter parametrizations in parallel and can be enriched with a foundational AI stack to allows for experiment comparison, sorter recommendation, and enhanced collaboration between and across researchers and teams.
Updates
Before DT4PH efforts, a NeuroCAAS was implemented, featuring
- functionality to upload recordings
- a web interface which
- provides an overview of experiments
- allows running a sorter algorithm with one or several sets of parameters
- a VM to perform curation and analyses on sorter outputs
We propose to enhance the existing NeuroCAAS platform with:
- UnitRefine: a Machine Learning model which predicts the quality of the sorting outcome and as such can drastically reduce the posthoc curation time, potentially even automate it, and facilitate comparison of several sorter outcomes
- a experimental graph neural net (GNN) based approach to detect neuronal activity as an alternative to the spike sorting / curation approach