Identification of Pattern Completion Neurons in Neuronal Ensembles Using Probabilistic Graphical Models.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Graph Theory

Background:

• Neuronal ensembles represent cognitive states but studying them is limited by tools to identify key neurons.
• Pattern completion neurons are crucial for recalling ensembles and guiding behavior, relevant to both biological and artificial neural networks.

Purpose of the Study:

• To develop and validate a method for reliably identifying and manipulating pattern completion neurons.
• To demonstrate the broad applicability and scalability of this method in neural circuit analysis.

Main Methods:

• Utilized conditional random fields (CRFs), a probabilistic graphical model, to identify pattern completion neurons.
• Applied CRFs to in vivo two-photon calcium imaging data from mouse visual cortex.
• Validated CRF predictions using two-photon optogenetics and analyzed public datasets and in silico simulations.

Main Results:

• CRFs successfully identified pattern completion neurons capable of activating entire neuronal ensembles in mice.
• The method reliably predicted neurons responding to specific visual stimuli in public datasets.
• In silico simulations showed CRFs-identified neurons possess increased functional connectivity.

Conclusions:

• Conditional random fields provide a powerful tool for characterizing and selectively manipulating neural circuits.
• This graph theory-based approach enables the identification of key neurons for ensemble recall and potential therapeutic applications.