Manuel Reyes-Sanchez,
Irene Elices,
Rodrigo Amaducci,
Francisco B Rodriguez,
Pablo VaronaVirtual Room
https://meet.google.com/ufe-frxk-mbyIn this work, we present an approach to automatically explore neuron and synapse model parameter to archive target dynamics or characterize emergent phenomena which rely on the temporal structure of biological recordings that are used as inputs to the models. The associated exploration and mapping allow us to assess the role of different elements in the equations of the neuron and synapse models to build a nontrivial integration of sequential information, which is also reflected in the time course of the corresponding model response. We illustrate this methodology in the context of dynamical invariants defined as cycle-by-cycle preserved relationships between time intervals that build robust sequences in neural rhythms. We have recently unveiled the existence of such invariants in the pyloric CPG of crustacean, even under the presence of intrinsic or induced large variability in the rhythms (Elices et al., 2019). The proposed strategy can be generalized for many types of neural recordings and models. During such protocol, we input biological data with a characteristic temporal structure to different model neurons. The biological recordings are preprocessed online to adapt the corresponding time and amplitude scales to those of the synapse and neuron models using a set of algorithms developed in our previous works (Amaducci et al., 2019; Reyes-Sanchez et al., 2020). Our methodology can then map the neuron and synapse parameters that yield a predefined dynamics taking into account the temporal structure of the model output. The algorithms allow for a full characterization of the parameter space that contributes to the generation of the predefined dynamics. To illustrate this protocol that combines experimental recordings and theoretical paradigms, we have applied it to the search for dynamical invariants established between a living CPG cell and a model neuron connected through a graded synapse model. Dynamical invariants are preserved cycle-by cycle, even during transients. In our validation tests, we have mapped the presence of a linear relationship, i.e. an invariant, between the interval defined by the beginning of the bursting activity of the two neurons (first- to-first spike interval between the living and model neurons) and the instantaneous period of their sequence in such hybrid circuit. The protocol has been used to assess the role of model and synaptic parameters in the generation of the dynamical invariant, achieving a high efficient mapping in a few minutes. We argue that this approach can also be employed to readily characterize optimal parameters in the construction of hybrid circuits built with living and artificial neurons and connections, and, generally, to validate neuron and synapse models.
Funded by AEI/FEDER PGC2018-095895-B-I00 and TIN2017-84452-R
ReferencesAmaducci, R., Reyes-Sanchez, M., Elices, I., Rodriguez, F. B., and Varona, P. (2019). RTHybrid: a standardized and open-source real-time software model library for experimental neuroscience. Front. Neuroinform. 13, 11.
doi:10.3389/fninf.2019.0001Elices, I., Levi, R., Arroyo, D., Rodriguez, F. B., and Varona, P. (2019). Robust dynamical invariants in sequential neural activity. Sci. Rep. 9, 9048.
doi:10.1038/s41598-019-44953-2Reyes-Sanchez, M., Amaducci, R., Elices, I., Rodriguez, F. B., and Varona, P. (2020).Automatic adaptation of model neurons and connections to build hybrid circuits with living networks. Neuroinformaticcs.
doi:10.1007/s12021-019-09440-z