Downloads: 102

Files in This Item:
File Description SizeFormat 
s12868-016-0255-x.pdf2.58 MBAdobe PDFView/Open
Title: Empirical Bayesian significance measure of neuronal spike response
Authors: Oba, Shigeyuki  kyouindb  KAKEN_id
Nakae, Ken  kyouindb  KAKEN_id
Ikegaya, Yuji
Aki, Shunsuke
Yoshimoto, Junichiro
Ishii, Shin  kyouindb  KAKEN_id
Author's alias: 大羽, 成征
中江, 健
石井, 信
Keywords: Empirical Bayesian testing
Functional connectivity of neurons
Generalized linear model
Spike statistics
Issue Date: 21-May-2016
Publisher: BioMed Central Ltd.
Journal title: BMC Neuroscience
Volume: 17
Thesis number: 27
Abstract: Background: Functional connectivity analyses of multiple neurons provide a powerful bottom-up approach to reveal functions of local neuronal circuits by using simultaneous recording of neuronal activity. A statistical methodology, generalized linear modeling (GLM) of the spike response function, is one of the most promising methodologies to reduce false link discoveries arising from pseudo-correlation based on common inputs. Although recent advancement of fluorescent imaging techniques has increased the number of simultaneously recoded neurons up to the hundreds or thousands, the amount of information per pair of neurons has not correspondingly increased, partly because of the instruments' limitations, and partly because the number of neuron pairs increase in a quadratic manner. Consequently, the estimation of GLM suffers from large statistical uncertainty caused by the shortage in effective information. Results: In this study, we propose a new combination of GLM and empirical Bayesian testing for the estimation of spike response functions that enables both conservative false discovery control and powerful functional connectivity detection. We compared our proposed method's performance with those of sparse estimation of GLM and classical Granger causality testing. Our method achieved high detection performance of functional connectivity with conservative estimation of false discovery rate and q values in case of information shortage due to short observation time. We also showed that empirical Bayesian testing on arbitrary statistics in place of likelihood-ratio statistics reduce the computational cost without decreasing the detection performance. When our proposed method was applied to a functional multi-neuron calcium imaging dataset from the rat hippocampal region, we found significant functional connections that are possibly mediated by AMPA and NMDA receptors. Conclusions: The proposed empirical Bayesian testing framework with GLM is promising especially when the amount of information per a neuron pair is small because of growing size of observed network.
Rights: © 2016 Oba et al. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (, which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
DOI(Published Version): 10.1186/s12868-016-0255-x
Appears in Collections:Journal Articles

Show full item record

Export to RefWorks

Export Format: 

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.