Paternal aging affects developmental convergence of vocal behavior in individual mice: machine learning-driven analyses of individuality

Abstract

AbstractChildren with autism spectrum disorder (ASD) show atypical developmental trajectories regarding a variety of phenotypes including vocal communication. Infant crying is a way of social communication and modeled in mice by recording ultrasonic vocalization (USV) induced by maternal separation. Here we applied a novel unsupervised machine learning approach (variational autoencoder, VAE), which can capture comprehensive features of USV to understand the complex phenotypes at an individual level in mice. First, the USVs from neonatal mice derived from Autism susceptibility candidate 2 (Auts2) gene mutant fathers were analyzed to validate the VAE. Then, we applied the VAE in a paternal aging model to trace developmental patterns of USV in pups born to aged fathers at the postnatal day 3 (P3), P6, P9, and P12. Our results revealed that pups derived from aged fathers emitted fewer USVs with less diversity and incomplete distribution patterns in an inferring map. While pups derived from young fathers showed a developmental convergence in vocal behavior, those from aged fathers exhibited more atypical patterns with wider variability. Thus, we demonstrate that paternal aging indeed has significant effects on early vocal development and subsequently increases the number of individuals with atypical behavior. The application of VAE also enables us to reveal individual difference, leading to the quantitative research of individuality.Significance StatementUltrasonic vocalization (USV) in mice is used for a model of vocal communication in autism spectrum disorder (ASD). We mathematically analyzed how maternal separation-induced USV was impaired in individual pups derived from aged fathers. Our novel approach with unsupervised machine learning (i.e., variational autoencoder, VAE), can detect key features of mouse USVs and identify individual differences during early postnatal stages. This study enables us to describe the individual difference, towards understanding individuality at the quantitative level. Individual variability is crucial in biomedical research to reveal new autistic-like features, investigate the first signs, stratify vocal communication phenotypes in animal models, and develop a translatable biological marker for diagnosing infant crying in the future.