Appears in collection : 2022 - T1 - WS2 - Mathematical modeling and statistical analysis in neuroscience

It was more than 50 years ago that Perkel, Gerstein, and Moore suggested a paradigm for detecting monosynaptic interactions from a cross-correlogram. While the original method may theoretically give plausible inferences, it sometimes suggests spurious connections in practice. This is because cross-correlograms are often accompanied by large fluctuations caused by common inputs to pairs of neurons. There have been many attempts to purge misinterpretations, by shuffling spike trains, by jittering spike times, or by taking fluctuating inputs into account. Here we developed a state-space method to detect the signature of monosynaptic connections or direct interaction buried in large fluctuations in a cross-correlogram. We evaluated the performance of the estimation accuracy by fitting the model to a large network of model neurons and then applied the model to real data.