Meeting Abstract
The small larval zebrafish (4mm) is capable of a large array of movements ranging from fast ones used to escape from predators, to the fine movements necessary for prey-capture. Usually, these movements are classified by scoring them into qualitatively defined discrete classes (e.g. J-turn, C-turn, etc.). However, recent work has shown that the speed and magnitude of swim bouts from these fish are graded. In the work presented here, we explore an alternative method of analysis with no a priori set of classes. First, we compute the net axial muscle forces along the fish body during swimming using a simplified body-fluid interaction model. Then, we investigate the inherent statistical structure of these muscle forces along the body from a set of over 650 swim bouts produced during free-swimming, prey capture, and the opto-motor reflex, a movement response in the direction of whole-field visual motion. We demonstrate that a linear combination of just 3 characteristic patterns of muscle forces along the body axis, or eigenforces, can explain over 99% of the variance seen in axial forces for all of the swim bouts. Since the time evolution of the coefficients of these eigenforces can define a bout, we use frequency and time domain characteristics of the coefficients to discover a small set of eigenbouts which, in linear combination, can recreate all of the bouts in the data set. These eigenbouts can be used to quantitatively define swim bouts in a lower order space and relate muscle activity to translation and rotation in the environment. We demonstrate that the eigenbouts are linearly related to total displacement and change in angular orientation. This work finds a small set of activity patterns which can quantitatively describe swim bouts and provide insight into how a variety of motor behaviors may be produced.
