<b>QUANTITATIVE BIOLOGY SEMINAR</B><br>Validity and Reliability of Granger Causality Analysis on Neuronal Network Reconstruction
Granger causality (GC) analysis is one of the major approaches to explore the dynamical causal connectivity among individual neurons or neuronal populations. In this talk, focusing on the connectivity reconstruction of the conductance-based integrate-and-fire neuronal networks, we address two issues: (i) how the causal connectivity obtained from GC analysis can be mapped to the underlying anatomical connectivity; and (ii) how we can sample discretely from the time continuous quantities, e.g., membrane potential, to obtain a reliable GC network reconstruction. We numerically demonstrate that the anatomical connectivity can be successfully reconstructed from the GC analysis and theoretically establish a quadratic relation between the GC and the coupling strength. We also analyze in detail the impact of sampling interval length on the GC analysis of uniformly sampled data and propose a strategy to circumvent the possible sampling hazards for a reliable network reconstruction. In addition, we establish a nonuniform sampling GC analysis framework to achieve a reliable network reconstruction. Finally, we note that our analysis on the validity and reliability of GC analysis can be extended to more general dynamics.