Studying sensory coding of complex naturalistic stimuli in the barrel cortex with a perturbative method and a machine learning approach – PerBaCo

Sensory perception relies on brain circuits that extract ethologically relevant features from the sensory organs. Understanding the feature “vocabulary” and how it can be used to build elaborate internal representations has been one of the main aims of sensory neuroscience. In particular, how sensory neurons in the primary somatosensory cortex, the brain area that receives and analyses tactile information, extract information non-linearly from complex, natural stimuli is still an open challenge. Most previous studies addressing this issue fall into two categories:
In a first category, physiological studies in anesthetized rodents examined what whisker movement features were encoded by artificially varying one stimulus parameter at a time. In a second category, a specific feature from natural stimuli is retained and used to test if sensory neurons are selective to that feature. While these approaches have uncovered the richness of sensory processing, they do not easily generalize to more complex stimuli.
The purpose of our project is to understand what features of the sensory flow are extracted by neurons in the barrel cortex - the area of the primary somatosensory cortex in rodents receiving information from the whiskers on the snout -, in the context of complex naturalistic stimuli. For this, we will design a novel approach, different from the two previous categories, where natural stimuli delivered to the whiskers are perturbed with noise while neuronal activity is recorded from the barrel cortex. Subsequently, a machine learning model is built to predict how barrel cortex neurons respond to these perturbations. A dimensionality reduction is applied to interpret the outcome of the model. This is the first time such a method will be applied to the somatosensory system in both anesthetized and awake head-posted mice. The combination of experiments and theory will give novel insights on the features encoded by the somatosensory cortex during natural stimulation.