Predicting how and when hidden neurons skew measured synaptic interactions

Area of Science:

• Computational neuroscience
• Systems neuroscience
• Network science

Background:

• Experimental neural recordings typically capture only a fraction of neurons, leading to skewed measurements.
• Interactions between recorded neurons and unobserved (
• hidden
• ) neurons complicate the interpretation of neural data.
• A deeper understanding of the relationship between measured and true neural properties is crucial for deciphering neural circuit function.

Purpose of the Study:

• To investigate how unobserved neurons alter the spatiotemporal dynamics of synaptic interactions.
• To develop a method for correcting measured neural properties based on the influence of hidden network components.
• To provide a quantitative framework for interpreting neural recordings in complex circuits.

Main Methods:

• Mathematical decomposition of effective synaptic interactions into direct and indirect pathways.
• Derivation of a formula quantifying the impact of unobserved neurons on observed neural interactions.
• Analysis of random networks with strong coupling (connection weights scaling with N).

Main Results:

• Effective neural interactions are a sum of direct interactions and contributions from all paths through unobserved neurons.
• A formula was derived to precisely quantify the influence of hidden units on measured synaptic interactions.
• The study identified conditions under which hidden units significantly reshape interactions among observed neurons.

Conclusions:

• The derived formula allows for the accurate interpretation of neural data by accounting for hidden network components.
• This work provides a quantitative relationship between measured and true neural interactions, crucial for understanding neural computation.
• The findings enable the study of how network properties shape effective interactions and how to potentially manipulate them.