Elements of a stochastic 3D prediction engine in larval zebrafish prey capture

Abstract

Many predatory animals rely on accurate sensory perception, predictive models, and precise pursuits to catch moving prey. Larval zebrafish intercept paramecia during their hunting behavior, but the precise trajectories of their prey have never been recorded in relation to fish movements in three dimensions.As a means of uncovering what a simple organism understands about its physical world, we have constructed a 3D-imaging setup to simultaneously record the behavior of larval zebrafish, as well as their moving prey, during hunting. We show that zebrafish robustly transform their 3D displacement and rotation according to the position of their prey while modulating both of these variables depending on prey velocity. This is true for both azimuth and altitude, but particulars of the hunting algorithm in the two planes are slightly different to accommodate an asymmetric strike zone. We show that the combination of position and velocity perception provides the fish with a preferred future positional estimate, indicating an ability to project trajectories forward in time. Using computational models, we show that this projection ability is critical for prey capture efficiency and success. Further, we demonstrate that fish use a graded stochasticity algorithm where the variance around the mean result of each swim scales with distance from the target. Notably, this strategy provides the animal with a considerable improvement over equivalent noise-free strategies.In sum, our quantitative and probabilistic modeling shows that zebrafish are equipped with a stochastic recursive algorithm that embodies an implicit predictive model of the world. This algorithm, built by a simple set of behavioral rules, allows the fish to optimize their hunting strategy in a naturalistic three-dimensional environment. INTRODUCTION:

Learning steers the ontogeny of an efficient hunting sequence in zebrafish larvae

“…The distributions of likely turn-ratios estimated by the model shown on Figure 5C confirm that the initial turns recorded in the LF group are distinct in undershooting prey azimuth; the slope of the regression model gives a mean 0.73 for LF, 0.87 NF, and 0.92 for DF. This undershooting behaviour recorded from the LF group is consistent with previous reports on the first-turn-to-prey of larvae that had also been reared with live prey (Paramecia) [8,43,51]. Paradoxically, hunt events from the NF and DF groups display initial turns that do not consistently undershoot, but rather align larvae closer to the prey's azimuth Figure 5C.…”

“…We focused our analysis of turn behaviour only on the initial turn towards prey because it amounts to the largest re-orientation following prey detection and so it can be representative of the ability to aim towards prey while being less prone to measurement error. Subsequent turn behaviour follows the same pattern of undershooting prey azimuth, and thus the first turn behaviour we observe is typical of a larva's re-orienting behaviour overall [8]. We hypothesize that this undershooting of the turns plays a role in enhancing capture success by improving the perception of prey distance.…”

“…We obtained further evidence that prey located off-axis is more likely to be captured, by observing that the distribution of initial prey azimuths across successful hunt episodes (see Figure S6 ) is bimodal. Although larvae are able to respond to prey in front of them equally well [6,8], we find that the distribution of prey azimuth is bimodal in successful hunt-events across all groups, and the maxima are located between 35°-50°on either side of the larval heading. Therefore, prey located off-axis at particular angles are more likely to be captured.…”

“…which assumes the errors are independent and identically distributed as normal random variables with mean zero and variance σ 2 , for which we used an inverse gamma prior σ 2 ∼ InvGamma(s, r), s = 5, r = 2 (8) while for the linear coefficients we used…”

Probabilistic Models of Larval Zebrafish Behavior Reveal Structure on Many Scales

“…These adjustments will allow for higher resolution pose estimation (e.g., by including pectoral fin dynamics, pitch and roll estimates, and tail half-beat analysis [13]), facilitate more comprehensive bout-type classification, and yield significantly longer continuous behavioral sequences. Richer datasets will enable future models to extract nuanced environmental dependencies, like prolonged attention to single prey among many distractors [65]. Future models may simultaneously infer discrete behavioral states and their dynamics [7,14,66,67], though the non-Markovian dependencies on past behavior present new challenges [68].…”