|Authors||M. E. Lepperød, T. Stöber, T. Hafting, M. Fyhn and K. Kording|
|Title||Inferring causal connectivity from pairwise recordings and optogenetics|
|Project(s)||Department of Computational Physiology|
|Publication Type||Journal Article|
|Year of Publication||2018|
|Publisher||Cold Spring Harbor Laboratory|
To study how the brain works, it is crucial to identify causal interactions between neurons, which is thought to require perturbations. However, when using optogenetics we typically perturb multiple neurons, producing a confound - any of the stimulated neurons can have affected the postsynaptic neuron. Here we show how this produces large biases, and how they can be reduced using the instrumental variable (IV) technique from econometrics. The interaction between stimulation and the absolute refractory period produces a weak, approximately random signal which can be exploited to estimate causal connectivity. When simulating integrate-and-fire neurons, we find that estimates from IV are better than naive techniques (R2=0.77 vs R2=0.01). The difference is important as the estimates disagree when applied to experimental data from stimulated neurons with recorded spiking activity. Presented is a robust analysis framework for mapping out network connectivity based on causal neuron interactions.