Analyse haut-débit du comportement spontané d'un organisme modèle « simple »

Abstract : The use of simple genetic model organisms has allowed large-scale investigation of cellular and molecular mechanisms in biology. The zebrafish larva (Danio rerio) is a simple vertebrate model with several advantages for the identification of the molecular and pharmacological basis of development and behavior thanks to genetic and chemical screens. Several experimental paradigms rely on a precise behavioral characterization associated with a genotype or a drug application. This phenotyping process would benefit from a global comprehension of all possible zebrafish larvae maneuvers in order to precisely characterize the global behavior of larvae observed during an experiment. The analysis over multiple minutes of the spontaneous locomotion of zebrafish larvae in a group provides an ideal framework for achieving these goals.The manual analysis of animal behavior is possible for the observation of stereotyped and isolated maneuvers, but it is not compatible with large-scale approaches. After five days, the zebrafish larva moves in burst of movement occurring at high frequency and separated by a few seconds. The difficulty of observation of high-frequency maneuvers makes a manual analysis of videos impossible. The recent development of high-speed cameras, coupled with advances in computer science and artificial intelligence, make an automatic analysis of behavior possible. The available commercial solutions allow long acquisitions but without capturing all the complexity and diversity of the movements. We thus created ZebraZoom, our high-throughput analysis system which makes it possible to follow, quantify and categorize the spontaneous behavior of all zebrafish larva in a group.We built an experimental setup which consisted in placing seven 5- to 7-day old zebrafish larvae in eight petri dishes. The petri dishes were placed on a light table and a high-speed camera was placed above the setup and allowed to record the spontaneous behavior at 337 Hz during four minutes. Once the video was acquired, ZebraZoom localized automatically each of the eight petri dishes and followed the head and the tail of each of the animals on the entire video, thanks to computer vision methods. ZebraZoom then identifies automatically all burst of movement.The behavior phenotyping was performed thanks to two complementary methods: the extraction of global parameters characterizing movement dynamics and the automatic categorization of movements into stereotyped maneuvers. We used global parameters analysis to characterize the effect of drugs acting on neurotransmitters receptors and to analyze the behavior of a blind mutant. The automatic categorization of movements was carried out thanks to machine learning procedures. We showed the utility of this categorization method in studying interactions between larvae at early stages. In conclusion, our ZebraZoom program makes it possible to carry out an automatic and complete phenotyping, and this approach could be used in other systems and experimental contexts.
Document type :
Complete list of metadatas

Cited literature [45 references]  Display  Hide  Download
Contributor : Abes Star <>
Submitted on : Friday, November 8, 2013 - 7:53:00 PM
Last modification on : Saturday, September 7, 2019 - 3:07:04 AM
Long-term archiving on : Sunday, February 9, 2014 - 10:35:14 AM


Version validated by the jury (STAR)


  • HAL Id : tel-00881755, version 1


Olivier Mirat. Analyse haut-débit du comportement spontané d'un organisme modèle « simple ». Sciences agricoles. Université René Descartes - Paris V, 2013. Français. ⟨NNT : 2013PA05T023⟩. ⟨tel-00881755⟩



Record views


Files downloads