Identification and Targeting of Cortical Ensembles

Abstract

Breaking the neural code requires the characterization of physiological and behavioral correlates of neuronal ensemble activity. To understand how the emergent properties of neuronal ensembles allow an internal representation of the external world, it is necessary to generate empirically grounded models that fully capture ensemble dynamics. We used machine learning techniques, often applied in big data pattern recognition, to identify and target cortical ensembles from mouse primary visual cortex in vivo leveraging recent developments in optical techniques that allowed the simultaneous recording and manipulation of neuronal ensembles with single-cell precision. Conditional random fields (CRFs) allowed us not only to identify cortical ensembles representing visual stimuli, but also to individually target neurons that are functionally key for pattern completion. These results represent the proof-of-principle that machine learning techniques could be used to design close-loop behavioral experiments that involve the precise manipulation of functional cortical ensembles.All rights reserved. No reuse allowed without permission.(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.

Controlling Visually Guided Behavior by Holographic Recalling of Cortical Ensembles

“…We then used singular value decomposition (SVD) to identify neuronal ensembles, statistically defined as significant vectors clusters (Carrillo-Reid et al, 2015a; Carrillo-Reid et al, 2015b). After the identification of the ensembles, we used a conditional random field (CRF) model (Figure 2C) to find neurons that were most representative for each ensemble, on the basis of their predictability and the node strength of functional connections between neurons (Figure 2D) (Carrillo-Reid et al, 2017a). These neurons, which have pattern completion capabilities, could be then targeted for two-photon optogenetic stimulation (Figure 2E) by using holographic spatial light modulator (SLM) microscopy (Nikolenko et al, 2008) for recalling of ensembles (Carrillo-Reid et al, 2017a).…”

“…To identify the neurons to be targeted by two-photon optogenetics we used conditional random fields (CRFs) to model the conditional probability distribution to see a given neuronal ensemble firing together (Carrillo-Reid et al, 2017a; Tang et al, 2016). We used CRFs to capture the contribution of specific neurons to the overall network activity defined by population vectors belonging to a given neuronal ensemble.…”

“…To explore this, we combined calcium imaging of neuronal populations (Yuste and Katz, 1991), two-photon microscopy (Denk et al, 1990; Yuste and Denk, 1995), and population analysis (Carrillo-Reid et al, 2017a; Carrillo-Reid et al, 2015a) to identify neuronal ensembles in primary visual cortex from awake mice performing a visually guided Go/No-Go behavioral task. Then, using two-photon holographic optogenetics (Nikolenko et al, 2008; Packer et al, 2015; Rickgauer et al, 2014; Yang et al, 2018), we activated specific sets of neurons during the presentation of visual stimuli to disrupt or recall cortical ensembles while measuring behavioral performance at different contrast levels of visual stimuli.…”

Two-Color Volumetric Imaging of Neuronal Activity of Cortical Columns

“…One advantage of our microscope is that we can image multiple cortical layers almost simultaneously, which enables us to define and study cortical ensembles across layers on the basis of the correlation structure of the population. Several computational approaches have been proposed for ensemble detection ( Avitan et al, 2017 ; Carrillo-Reid et al, 2015 , 2017b ; Lopes-dos-Santos et al, 2013 ); because we record from a relatively large population of neurons, we chose to use a fast graph-based community detection method (the Louvain method; Blondel et al, 2008 ), whose aim is to maximize modularity measurement. To detect stable ensembles, we combined the Louvain method with consensus clustering, which finds the best agreement between repetitions ( Lancichinetti and Fortunato, 2012 ).…”

Dual-color volumetric imaging of neural activity of cortical columns

“…One advantage of our microscope is that we can image multiple cortical layers almost simultaneously, which enables us to define and study cortical ensembles across layers based on the correlation structure of the population. Several computational approaches have been proposed for ensemble detection (Avitan et al, 2017;Carrillo-Reid et al, 2015, 2017bLopes-dos-Santos et al, 2013); since we record from a relatively large population of neurons, we chose to use a fast graph-based community detection method [Louvain method (Blondel et al, 2008)] which aims at maximizing modularity measurement. To detect stable ensemble, we combined the Louvain method with consensus clustering that finds the best agreement between repetitions (Lancichinetti and Fortunato, 2012).…”