Toward automatic behavioral screen: a computational model for analyzing Caenorhabditis elegans locomotion

Abstract

Genetic screen has been facilitating molecular geneticists to analyze mutants that produce certain phenotypes. However traditional methods for behavioral phenotype screen of mutant Caenorhabditis elegans rely on human observers and therefore are subjective and imprecise. This work dedicates a model to quantify and analyze the worm behavior using automatically-tracking and time-coded images. We have delved into the following questions: (1) how to achieve simplified worm-shape representation, (2) how to describe worm locomotion, (3) how to obtain frequent locomotion patterns and then representative behavioral patterns, and (4) how to discover interesting behavioral actions within a representative behavioral pattern. Since the methodologies focus on rigorous image-based behavioral screening and phenotyping, the proposed methods should be trustworthy for behavior analysis of tiny organisms based on their microscopic video frames.