A graph network model for neural connection prediction and connection strength estimation. (1st June 2022)

• Record Type: Journal Article
• Authors: Yuan, Ye; Liu, Jian; Zhao, Peng; Wang, Wei; Gu, Xiao; Rong, Yi; Lai, Tinggeng; Chen, Yuze; Xin, Kuankuan; Niu, Xin; Xiang, Fengtao; Huo, Hong; Li, Zhaoyu; Fang, Tao
• Is Part Of: Journal of neural engineering. Volume 19:Number 3(2022)
• Publication Date: 2022-06-01
• Subjects: C. elegans modeling -- graph neural network -- neural connection presiction -- synaptic strength estimation -- C. elegans connectome; Neurosciences -- Periodicals; Biomedical engineering -- Periodicals; 612.8
• DOI: 10.1088/1741-2552/ac69bd
• Languages: English
• ISSNs: 1741-2560

Abstract

Abstract: Objective . Reconstruction of connectomes at the cellular scale is a prerequisite for understanding the principles of neural circuits. However, due to methodological limits, scientists have reconstructed the connectomes of only a few organisms such as C. elegans, and estimated synaptic strength indirectly according to their size and number. Approach . Here, we propose a graph network model to predict synaptic connections and estimate synaptic strength by using the calcium activity data from C. elegans. Main results . The results show that this model can reliably predict synaptic connections in the neural circuits of C. elegans, and estimate their synaptic strength, which is an intricate and comprehensive reflection of multiple factors such as synaptic type and size, neurotransmitter and receptor type, and even activity dependence. In addition, the excitability or inhibition of synapses can be identified by this model. We also found that chemical synaptic strength is almost linearly positively correlated to electrical synaptic strength, and the influence of one neuron on another is non-linearly correlated with the number between them. This reflects the intrinsic interaction between electrical and chemical synapses. Significance . Our model is expected to provide a more accessible quantitative and data-driven approach for the reconstruction of connectomes in more complex nervous systems, as well as a promising method for accurately estimating synaptic strength.