Robust characterization of selectivity of individual neurons to distinct task-relevant behavioral states using calcium imaging

Abstract

Investigations into the neural basis of behavior have recently employed fluorescence imaging of calcium dynamics in a variety of brain areas to measure neural responses. However, across studies, diverse and seemingly subjective methodological choices have been made in assessing the selectivity of individual neurons to task-relevant behavioral states. Here, we examine systematically the effect of different choices in the values of key parameters from data acquisition through statistical testing on the inference of the selectivity of individual neurons for task states. We do so by using as an experimental testbed, neuronal calcium dynamics imaged in the medial prefrontal cortex of freely behaving mice engaged in a classic exploration-avoidance task involving spontaneous (animal-controlled) state transitions - navigation in the elevated zero maze (EZM). We report that a number of key variables in this pipeline substantially impact the selectivity label assigned to neurons, and do so in distinct ways. By quantitatively comparing newly defined accuracy and robustness metrics for all the 128 possible combinations of levels of the key parameters, we discover in a data-driven manner, two optimal combinations that reliably characterize neuronal selectivity - one using discrete calcium events and another using continuous calcium traces. This work establishes objective and standardized parameter settings for reliable, calcium imaging-based investigations into the neural encoding of task-states.