A mixture distribution approach for assessing genetic impact from twin study

2025-07-17

AI Summary

In twin studies, conventional methods struggle to accurately estimate heritability of immune traits when the within-pair birth order of twins is unknown. To address this, we propose a joint-likelihood-based mixture binary distribution model that directly models unordered monozygotic and dizygotic twin pairs, thereby circumventing arbitrary ordering assumptions and avoiding information loss. By constructing an identifiable joint likelihood function, our approach achieves √n-consistent estimation, markedly improving convergence speed and statistical efficiency. This method overcomes key limitations of classical mixture models—namely, slow convergence and nonidentifiability—when applied to unordered paired data, while delivering high estimation accuracy and robustness. The resulting framework provides a general, reliable, and computationally efficient statistical methodology for heritability inference of complex phenotypes such as immune traits, and is readily extendable to other settings involving unordered paired data.

Technology Category

• Search and Optimization: Mixed Discrete/Continuous Search
• Reasoning under Uncertainty: Relational Probabilistic Models
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

This work was motivated by a twin study with the goal of assessing the genetic control of immune traits. We propose a mixture bivariate distribution to model twin data where the underlying order within a pair is unclear. Though estimation from mixture distribution is usually subject to low convergence rate, the combined likelihood, which is constructed over monozygotic and dizygotic twins combined, reaches root-n consistency and allows effective statistical inference on the genetic impact. The method is applicable to general unordered pairs.

Problem

• Assessing genetic control of immune traits in twins
• Modeling twin data with unclear pair order using mixtures
• Improving convergence in genetic impact estimation for pairs

Innovation

• Mixture bivariate distribution models twin data
• Combined likelihood ensures root-n consistency
• Applicable to general unordered pairs