Deep learning of mother-infant interactions in V1b Vasopressin receptor knockout mice

Abstract

【Background】Oxytocin, a neurohypophysial hormone from the posterior pituitary, is known as an important factor for childcare and breastfeeding. Also, vasopressin has been reportedly involved in maternal behaviors through the vasopressin receptors of V1A and V1B subtypes. Previous studies demonstrated that the V1A receptor antagonist given into the median preoptic area of rat resulted in significant reduction of caregiving behavior of the mother. Although the studies on the oxytocin hormone and vasopressin/V1A receptor have been extensively conducted, our knowledge on the V1B receptor in maternal behavior is still limited.【Purpose】We intended to clarify a role of the V1B receptor in mother-child interaction during lactating period.【Methods】We compared exploratory behavior between nonpregnant females of control mice and those of the V1B knockout mice in open field test. After giving a birth, the mothers were examined on their behavior in pup retrieval test. Moreover, massive amounts of data from behavioral recordings were visualized and mother-infant relationship was analyzed by deep learning strategy.【Results and discussion】After training about 3000 images, our deep learning model successfully classified mother and babies with 99% accuracy. The analysis results by deep learning model were in good agreement with the observational results by an investigator. Together, we propose that this new method can be applied further to other areas of behavioral study to overcome the limitations and increase the efficiency for analyzing complex behaviors.