580 California St., Suite 400
San Francisco, CA, 94104
Key research themes
1. How can multivariate linear mixed models be efficiently applied to bivariate genetic analysis to improve the detection of genetic components across correlated traits?
This theme focuses on the application and development of multivariate linear mixed models (LMMs) tailored to bivariate or multivariate genetic data. The goal is to accurately partition variance components attributable to shared genetic effects between traits, while accounting for population structure and confounding factors. Efficient computational implementations and model flexibility are critical to handle large datasets and multiple traits simultaneously, thereby enhancing power to detect genetic associations and improving prediction accuracy.
Key finding: Introduces LIMIX, a computationally efficient and flexible multi-trait mixed modeling framework enabling joint analysis of tens to hundreds of phenotypes. LIMIX allows modeling of different fixed and random effects across... Read more
Key finding: Demonstrates that average semivariance (ASV) can be incorporated within a linear mixed model framework to simultaneously estimate genetic variance components for Mendelian, oligogenic, and polygenic terms, including additive... Read more
Key finding: Derives and validates a deterministic equation predicting genomic prediction accuracy when data from multiple traits and populations are combined, incorporating genetic correlations between traits and populations, trait... Read more
2. What statistical methods can robustly account for non-normality and discrete phenotypes in bivariate genetic association studies?
This research theme addresses methodological challenges in bivariate genetic analyses involving discrete, binary, or mixed-type traits where traditional multivariate LMM assumptions of multivariate normality do not hold. Developing flexible statistical approaches, such as copula models, to handle mixed phenotype types improves the applicability of bivariate genetic tests to complex disorders and traits with non-continuous measures, increasing power and interpretability of genetic associations.
Key finding: Proposes CBMAT, a unified copula-based method for region-based genetics association testing accommodating bivariate continuous, binary, or mixed phenotypes. By modeling non-normal and discrete traits jointly through copula... Read more
Key finding: Introduces a Bayesian group test approach equivalent to a linear mixed model with two random effects to jointly test sets of genetic variants for association with traits, explicitly modeling confounding due to population... Read more
3. How can locus heterogeneity and complex inheritance patterns be uncovered and modeled in bivariate genetic studies to enhance gene identification?
This theme investigates statistical and genetic challenges posed by familial locus heterogeneity and polysomic inheritance in bivariate or multivariate genetic analyses. The presence of locus heterogeneity can confound genetic mapping, while complex inheritance modes (e.g., polysomic polyploidy) require tailored parentage and genetic variance modeling. Methods capable of detecting, accounting for, and modeling these complexities improve gene discovery success rates and clarifies genetic architecture underlying bivariate traits.
Key finding: Demonstrates that familial locus heterogeneity, where different causal variants segregate within a single pedigree, complicates gene identification in Mendelian traits. They utilize linkage analysis and homozygosity mapping... Read more
Key finding: Develops a new parentage analysis model accounting for polysomic inheritance, double reduction, null alleles, genotyping errors, and self-fertilization, overcoming challenges of genotypic ambiguity in polyploids. The model... Read more
Key finding: Evaluates the power and false positive rates of epistatic QTL mapping methods and finds that standard approaches maximize QTL detection but suffer from high false positives, especially for epistatic interactions. The results... Read more