To Simulate Neural Organoid: A Framework and A Benchmark based on AI

Jingjie Shang, Haorui Zheng, Chengyi Zhang, Jian Cao, Liqun Chen, Xiaochen Bo, Yuan Wang

Primary Area: applications to neuroscience & cognitive science

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Keywords: Neural Organoid Simulation, Spiking Neural Network, AI for neurosciences

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

Abstract:

Neural organoids hold significant value for AI modeling, cognitive exploration, and medical discovery in academic research.
However, the current research on neural organoids primarily relies on trial-and-error experimental design methods that are time-consuming and expensive.
Moreover, due to the intrinsic unknowability of complex biological systems, purely rational deduction through mathematical and physical modeling is nearly impossible.
As a result, the design of organoid experiments is constrained by the above limitations.
With AI models being applied to address diverse biological challenges, the demand for novel experimental paths for neural organoids has become urgent.
In response to the above issues, we propose the first neural organoid simulation framework to realistically reconstruct various details of interaction experiments using real mature organoids.
This framework employs advanced neural computing models as elements, harnessing AI methods to enable stimulation, response, and learning functionalities.
The significant consumption can be mitigated through the combination of the framework with real experiments.
An intelligent expansion platform is also established based on spiking neural network to facilitate the exploration of organoid-machine collaborative intelligence.
In addition, we introduce a benchmark for evaluating our framework, including a set of real organoid experimental data and a series of evaluation metrics.
The experimental results show that our simulation framework features outstanding simulation capabilities and reflects similarity with real organoid experiments in many aspects.
With the intelligent expansion platform, the performance of the combination is comparable to pure AI algorithms in a basic classification task.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 787