Analysis of Spatio-Temporal Spike Patterns in experimental parallel recordings with SPADE

Correlated neuronal activity is considered to be the signature of an active cell assembly, i.e. a group of neurons assumedto comprise the building blocks of information processing in the brain (Hebb, 1949). We suppose that activations of aparticular cell assembly are expressed by the repetitive occurrences of precise spatio-temporal spike patterns (STP)composed of member neurons of the cell assembly. In order to detect these patterns, we employ a method called SpikePAttern Detection and Evaluation (SPADE; Quaglio et al. 2017). The method can cope with massively parallel spikedata, by first counting repeating STPs using a frequent itemset approach (Torre et al, 2013), followed by a significanceevaluation. To avoid massive multiple testing, patterns of the same signature, i.e. number of spikes involved, number ofoccurrences and pattern duration, are pooled. The significance of each signature is evaluated by comparing the numberof patterns of that signature to surrogate data.We aim to test if a particular assembly is activated in the motor cortex when a particular movement is performed. Fordoing so, we apply SPADE to simultaneous spike data recorded by a 10x10 Utah array from pre-/motor cortex of twomacaque monkeys (Riehle et al. 2013). The animals performed a delayed reach to grasp task. After a preparatoryperiod, they were instructed to pull and hold an object by two possible grips (side grip SG, or precision grip PG) andwith two possible force levels (high force HF, or low force LF), giving thus four behavioural conditions. We analyzeone session (of 15min) of each of the two monkeys (data publicly available, see Brochier et al, 2018). Each trial type(e.g. SG-HF) is segmented into six epochs of 500ms, each of which are analyzed separately. We find in both monkeysthat a) most STPs are found during the movement period; b) all patterns found in this period are composed of differentcompositions of the same few neurons; c) the pattern compositions vary in the different movement behaviors; d) withinone movement condition we find one neuron, that is involved in all the patterns, and e) that its individual spikesparticipate multiply in different patterns. The latter results suggests to consider these neurons as hub neurons.Further analysis will include several sessions of the same experiment for each monkey, in order to see if theseobservations are of general nature