Biological systems are in a constant state of flux and neuronal networks are dynamic, new neurons are added to circuits during development and during adulthood synaptic plasticity allows for formation of new, temporary connections and removal of the ones no longer needed. Understanding how the brain processes information is a challenging task that must add a constantly changing network, a dynamic microenvironment, and the continuous information inflow from the outside environment, to the already complex equation. Dr. Tatyana Sharpee at the Salk Institute conducts groundbreaking research integrating physics, mathematics and information theory with neural computations in order to understand how the brain processes sensory information. By understanding data from experiments, her lab develops new mathematical and statistical frameworks aimed to explain sensory processing and predict animal behavior. By using information theory her research seeks to understand how communication happens across different areas in the brain. The disruption in information processing leads to disease. Understanding how the information is processed and what leads to its disruption could help to develop brain-machine interfaces which could be used therapeutically in the future. Dr. Sharpee has taken different approaches to understand how the brain functions as a biological system, from object recognition involving visual stimuli to olfaction.
One approach the field has used in understanding how the sum of a complex neuronal network yields a functional operation is termed neuronal population vector. The neuronal population vector is a weighted vectorial sum of individual elements, neurons, and their activity and results in an estimated measure of behavior. Although this approach has advanced the field, it has some drawbacks such as discarding substantial information included in the responses of a neuronal population. This limits the understanding of how signal communication occurs between different areas within the central nervous system. By using information theory, Dr. Sharpee and her team have modified the population vector expression to achieve a blueprint for building circuits where signals can be read-out without information loss, an approach they have named sufficient population vector .
The sufficient vector captures all information from diverse neural populations and with correlated variability across neurons. In all panels, they compare information transmitted by a population response (black line) with information transmitted by the sufficient population vector (red) and the standard population vector (dashed gray). Neural populations tuned to the same (a, c) or different (b, d) preferred stimuli. In (a) differences in neural tuning curves are due to differences in steepness values, whereas in (c) they are due to differences in thresholds. Dotted lines show the information values obtained by binning response variables. Dotted lines overlap with solid curves. (d) same as (b) but with noise correlations. Insets show example population tuning curves. Taken from Sharpee et. al. 2019.
To hear more about Dr. Sharpee’s sufficient population vector and recent projects, join us at her talk titled “Reading responses of large neural population without information loss” Tuesday April 21, 2020 at 4 pm via Zoom (https://uchealth.zoom.us/j/501283195).
Written by Minerva Contreras a 1st year in the Neurosciences Graduate Program at UCSD.
 Mahan, M. Y., & Georgopoulos, A. P. (2014). Neuronal Population Vector. In Encyclopedia of Computational Neuroscience (pp. 1–7). Springer New York. https://doi.org/10.1007/978-1-4614-7320-6_401-1
 Sharpee, Tatyana O., & Berkowitz, John A. (2019). Linking neural responses to behavior with information-preserving population vectors. Current Opinion in Behavioral Sciences, 29(C). doi:10.1016/j.cobeha