Figure 6 : Behavioural characterization of four mouse strains.

From: Automated home-cage behavioural phenotyping of mice

Figure 6

(a) The average total resting time for each of the four strains of mice over 24 h (n=7 animals for each strain). (b) Average duration of resting bouts (defined as a continuous duration with one single behaviour). Values of mean ± s.e.m. are shown, *P<0.01 by analysis of variance with Tukey's post hoc test. (c) Total time spent grooming exhibited by the BTBR strain as compared with the C57BL/6J strain within 10–20 min after placing the animals in a novel cage. Values of mean±s.e.m. are shown, *P<0.05 by Student's t-test, one-tailed. P=0.04 for system and P=0.0254 for human 'H', P=0.0273 for human 'A'). (d, e) Characterizing the genotype of individual animals on the basis of the patterns of behaviour measured by the computer system. The pattern of behaviours for each animal is a 32-dimensional vector, corresponding to the relative frequency of each of the eight behaviours of interest, as predicted by the system, over a 24-h period. (d) Principal component analysis to visualize the similarities/dissimilarities between patterns of behaviours exhibited by all 28 individual animals (seven mice×four strains). (e) Confusion matrix for a Support Vector Machine (SVM) classifier trained on the patterns of behaviour using a leave-one-animal-out procedure. The SVM classifier is able to predict the genotype of individual animals with an accuracy of 90% (chance level is 25% for this four-class classification problem). The confusion matrix shown here indicates the probability for an input strain (along the rows) to be classified, on the basis of its pattern of behaviour, as each of the four alternative strains (along the columns). The higher probabilities along the diagonal and the lower off-diagonal values indicate successful classification for all categories. For example, the value of 1.0 for the C57BL/6J strain means that all C57BL/6J animals were correctly classified as such.