Bayesian Semiparametric Markov Renewal Mixed Models for Vocalization Syntax

TL;DR

This paper introduces a novel Bayesian semiparametric Markov renewal mixed model to analyze mouse vocalization data, capturing both transition probabilities and inter-syllable interval times with covariate and individual effects.

Contribution

It develops a flexible modeling framework that jointly analyzes transition dynamics and interval times, incorporating covariates and random effects, which was lacking in prior models.

Findings

Genotype and social context significantly influence inter-syllable interval times.

Genotype and social context have weaker effects on transition dynamics than previously thought.

The model provides new insights into the vocal behavior affected by the Foxp2 mutation.

Abstract

Studying the neurological, genetic and evolutionary basis of human vocal communication mechanisms is an important field of neuroscience. In the absence of high quality data on humans, mouse vocalization experiments in laboratory settings have been proven to be useful in providing valuable insights into mammalian vocal development and evolution, including especially the impact of certain genetic mutations. Data sets from mouse vocalization experiments usually consist of categorical syllable sequences along with continuous inter-syllable interval times for mice of different genotypes vocalizing under various contexts. Few statistical models have considered the inference for both transition probabilities and inter-state intervals. The latter is of particular importance as increased inter-state intervals can be an indication of possible vocal impairment. In this paper, we propose a class of novel Markov renewal mixed models that capture the stochastic dynamics of both state transitions and inter-state interval times. Specifically, we model the transition dynamics and the inter-state intervals using Dirichlet and gamma mixtures, respectively, allowing the mixture probabilities in both cases to vary flexibly with fixed covariate effects as well as random individual-specific effects. We apply our model to analyze the impact of a mutation in the Foxp2 gene on mouse vocal behavior. We find that genotypes and social contexts significantly affect the inter-state interval times but, compared to previous analyses, the influences of genotype and social context on the syllable transition dynamics are weaker.