Measuring the amount of computation done in C.elegans brain dynamics

Biological systems ranging from genetic circuits to the human brain perform computations. Although extensive research has been done on the functional and behavioral traits accompanying these computations, the equally crucial question of quantifying the precise computation done in biological systems, or at a minimum, the amount of computation done, is under-investigated. Here we introduce a novel approach, and use it to quantify the amount of computation done in the joint dynamics of the neurons in a C.elegans. Our approach uses recent breakthroughs in neural stochastic differential equations to estimate the requisite latent dimensionality of the joint dynamics of the neurons, based on time series data sets of that dynamics. In turn, our time series data are found using fluorescent microscopy to acquire the whole brain activity of C.elegans. Our approach finds that high and low amounts of computation are required in, respectively, (the joint neuronal dynamics of) freely moving and immobile worms. Our approach also provides insights into the amount of computation occurring during other behavioral states of the worm, e.g., when it is swimming backwards.