A Probabilistic Framework for Decoding Behavior From in vivo Calcium Imaging Data

Abstract

Understanding the role of neuronal activity in cognition and behavior is a key question in neuroscience. Previously, in vivo studies have typically inferred behavior from electrophysiological data using probabilistic approaches including Bayesian decoding. While providing useful information on the role of neuronal subcircuits, electrophysiological approaches are often limited in the maximum number of recorded neurons as well as their ability to reliably identify neurons over time. This can be particularly problematic when trying to decode behaviors that rely on large neuronal assemblies or rely on temporal mechanisms, such as a learning task over the course of several days. Calcium imaging of genetically encoded calcium indicators has overcome these two issues. Unfortunately, because calcium transients only indirectly reflect spiking activity and calcium imaging is often performed at lower sampling frequencies, this approach suffers from uncertainty in exact spike timing and thus activity frequency, making rate-based decoding approaches used in electrophysiological recordings difficult to apply to calcium imaging data. Here we describe a probabilistic framework that can be used to robustly infer behavior from calcium imaging recordings and relies on a simplified implementation of a naive Baysian classifier. Our method discriminates between periods of activity and periods of inactivity to compute probability density functions (likelihood and posterior), significance and confidence interval, as well as mutual information. We next devise a simple method to decode behavior using these probability density functions and propose metrics to quantify decoding accuracy. Finally, we show that neuronal activity can be predicted from behavior, and that the accuracy of such reconstructions can guide the understanding of relationships that may exist between behavioral states and neuronal activity.

Conjunctive spatial and idiothetic codes are topographically organized in the GABAergic cells of the lateral septum

“…We subsequently filtered and binarized the recorded calcium traces (Fig. 3b) 47 and excluded periods of immobility (< 5cm.s −1 ). We found that LS GABAergic activity was significantly spatially modulated (Fig.…”

“…3f). Using an information-theoretic framework 48–50 , we computed the mutual information (MI) between calcium activity and spatial location for each cell and expressed the results in bits 47 . To circumvent the inherent issue of sparse sampling in 15min long recordings, we computed the mean MI and confidence interval using 30 bootstrap samples (50% data sampling per sample, Fig.…”

“…We recorded Ca 2+ activity of 1082 LS neurons (N = 15 mice), 1148 dCA1 pyramidal cells (N = 5 mice) and 144 cells from dCA3 (N = 2 mice) from animals alternating on a linear track. We subsequently filtered and binarized the recorded calcium traces (Supplemental figure 2A, Etter et al, 2020) and excluded periods of immobility (< 5cm.s -1 ). Using an information-theoretic framework (Shannon, 1948; Skaggs et al, 1993; Souza et al, 2018), we computed the mutual information (MI) between calcium activity and spatial location for each cell and expressed the results in bits (30 bootstrap samples, 50% data sampling, Supplemental figure 2B-C, Etter et al, 2020).…”

“…We subsequently filtered and binarized the recorded calcium traces (Supplemental figure 2A, Etter et al, 2020) and excluded periods of immobility (< 5cm.s -1 ). Using an information-theoretic framework (Shannon, 1948; Skaggs et al, 1993; Souza et al, 2018), we computed the mutual information (MI) between calcium activity and spatial location for each cell and expressed the results in bits (30 bootstrap samples, 50% data sampling, Supplemental figure 2B-C, Etter et al, 2020). We found that LS GABAergic activity was significantly spatially modulated in comparison to circularly shuffled surrogates ( p < .01, Figure 1D, Supplemental figure 2C, H).…”

Calcium Imaging and the Curse of Negativity

“…The NMF approach, when applied on extracted neuronal activity data normalized with z-scoring or ΔF/F 0 , discards negative deviations from the baseline fluorescent signal. Another approach that we and others have used, the binarization of the data based on a threshold of activity to generate “bar codes” of the brain activity, also has an intrinsic non-negative assumption (Kubo et al, 2014 ; Naumann et al, 2016 ; Heap et al, 2018 ; Daviu et al, 2020 ; Etter et al, 2020 ). Other threshold-based approaches, or even data cleaning steps, run the risk of discarding all negative deviations from baseline, biasing conclusions drawn from the dataset to exclude inhibition from the modeled system.…”