Model-based detection of putative synaptic connections from spike recordings with latency and type constraints

Abstract

8Detecting synaptic connections using large-scale extracellular spike recordings presents a statistical challenge. 9While previous methods often treat the detection of each putative connection as a separate hypothesis test, here 10 we develop a modeling approach that infers synaptic connections while incorporating circuit properties learned 11 from the whole network. We use an extension of the Generalized Linear Model framework to describe the cross-12 correlograms between pairs of neurons and separate correlograms into two parts: a slowly varying effect due to 13 background fluctuations and a fast, transient effect due to the synapse. We then use the observations from all 14 putative connections in the recording to estimate two network properties: the presynaptic neuron type (excitatory 15 or inhibitory) and the relationship between synaptic latency and distance between neurons. Constraining the 16 presynaptic neuron's type, synaptic latencies, and time constants improves synapse detection. In data from 17 simulated networks, this model outperforms two previously developed synapse detection methods, especially on 18 the weak connections. We also apply our model to in vitro multielectrode array recordings from mouse 19 somatosensory cortex. Here our model automatically recovers plausible connections from hundreds of neurons, 20 and the properties of the putative connections are largely consistent with previous research. 21 22 Using in vivo or in vitro multielectrode arrays, the extracellular spiking of hundreds of neurons can be recorded 24 simultaneously. These recordings are allowing new, large-scale studies of neuronal networks (Hahn et al. 2019; 25 Harris et al. 2003; Levenstein et al. 2019; Okun et al. 2015; Tingley and Buzsáki 2018), and the number of 26 neurons that can be simultaneously recorded is increasing approximately exponentially (Stevenson and Kording 27 2011). Depending on the species, brain area, and electrode configuration, these simultaneously recorded 28 neurons can have tens of thousands of potential synapses between them. Detecting and characterizing these 29 synapses represents a major challenge for neural data analysis. Here, we develop a model-based method 30 incorporating network-level constraints on 1) the presynaptic neuron type and 2) the synaptic latencies between 31 pre-and postsynaptic neurons. We examine whether these constraints can improve synapse detection using 32 simulated data and large-scale in vitro multielectrode array recordings. 33 34Detecting synaptic connections from extracellular spike observations is a difficult statistical problem. Since both 35 spiking and synapses themselves are sparse, it is often difficult to distinguish between changes in spike 36 probability that are due to a specific synaptic input, changes that are due other (typically unobserved) inputs, or 37 due to chance. Using extracellular spike data, researchers often identify putative monosynaptic connections by 38 examining cross-correlograms between the spiking of two neurons. If two neu...

Predictable fluctuations in excitatory synaptic strength due to natural variation in presynaptic firing rate

“…For each pair of neurons that satisfies the testing criteria we then fit a parametric model that aims to estimate the latency, timescale, and strength of a putative synaptic effect. Here we fit an extended GLM based on the model used in Ren et al, (2020). Briefly, we describe the CCG using two components: 1) a slow fluctuation due to background fluctuations and 2) a fast, transient effect due to the synaptic effect.…”

“…Unlike intracellular signals that contain postsynaptic responses to single presynaptic spikes, extracellular spikes are sparse binary events, and studies of synapses often rely on the cross-correlograms (CCGs) between the pre- and postsynaptic spike trains. If two neurons are connected with an excitatory synapse, there is often a fast-onset, short-latency peak in the CCG (Perkel et al, 1967; Fetz et al, 1991), and several techniques have been developed to automatically detect these connected pairs (Barthó et al, 2004; Amarasingham et al, 2012; Kobayashi et al, 2019; Ren et al, 2020).…”

Inhibition-Dominated Rich-Club Shapes Dynamics in Cortical Microcircuits

“…To examine the cell-type composition of these distinct subnetworks we used spike waveform features (trough-to-peak duration and half amplitude duration) and firing rates to classify neurons as putative excitatory or inhibitory neurons (Ren et al, 2020) (see Methods) (Figure 1C). Using this classification scheme, we were able to assign putative cell types to 97% of the recorded neurons and the different cell types formed clearly separable clusters in feature space (Figure 1D) with distinct firing rate distributions for both cell types (Figure 1E).…”

A convolutional neural network for estimating synaptic connectivity from spike trains

“…There were many possible sources for the lack of reliability and specificity, such as large fluctuations produced by external signals or higher-order interactions among neurons. Over the years, there have been many attempts to minimize the presence of such spurious connections, by shuffling spike trains 2 , by jittering spike times [3][4][5][6] , or by taking fluctuating inputs into account [7][8][9][10][11][12][13] . These, in general, helped eliminate the FPs, but they then tended to be conservative, giving rise to false negatives (FNs), i.e., missing existing connections.Recently, we developed an estimation method by applying the generalized linear model (GLM) to each cross-correlogram 14 .…”

“…There are many algorithms that were developed to estimate synaptic connections from spike trains [1][2][3][4][5][6][7][8][9][10][11][12][13]19,[29][30][31] . We compared CoNNECT with the conventional cross-correlation method (CC) 19 , Jittering method 4 , Extended GLM 13 , and GLMCC 21 for their ability to estimate connectivity using synthetic data. Figure 3 shows connection matrices determined by the four methods referenced to the true connection matrices.…”