Beyond Static Brain Atlases: AI-Powered Open Databasing and Dynamic Mining of Brain-Wide Neuron Morphometry
Published in BioRxiv, 2024
Recommended citation: Jiang, S., Wang, L., Yun, Z., Chen, H., Yao, J., & Peng, H. (2024). Beyond Static Brain Atlases: AI-Powered Open Databasing and Dynamic Mining of Brain-Wide Neuron Morphometry (p. 2024.09.22.614319). bioRxiv. https://doi.org/10.1101/2024.09.22.614319 https://www.biorxiv.org/content/10.1101/2024.09.22.614319v1
Abstract:
We introduce NeuroXiv (neuroxiv.org), a large-scale, AI-powered database that provides detailed 3D morphologies of individual neurons mapped to a standard brain atlas, designed to support a wide array of dynamic, interactive neuroscience applications. NeuroXiv offers a comprehensive collection of 175,149 atlas-oriented reconstructed morphologies of individual neurons derived from more than 518 mouse brains, classified into 292 distinct types and mapped into the Common Coordinate Framework Version 3 (CCFv3). Different from conventional static brain atlases that are often limited to data-browsing, NeuroXiv allows interactive analyses as well as uploading and databasing custom neuron morphologies, which are mapped to the brain atlas for objective comparisons. Powered by a cutting-edge AI engine (AIPOM), NeuroXiv enables dynamic, user-specific analysis and data mining. We specifically developed a mixture-of-experts algorithm to harness the capabilities of multiple large language models. We also developed a client program to achieve more than 10 times better performance compared to a typical server-side setup. We demonstrate NeuroXiv’s scalability, efficiency, flexibility, openness, and robustness through various applications.