Canada Spends Logo

Explorez notre analyse du budget 2026

Grant: Discovery Grants Program - Individual

University of Toronto

$44,000.00

Novel robust methods for complex binary traits

EF 2024-25

Résumé
In the last two decades, many advanced statistical methods have been proposed, developed and implemented for genetic studies of human complex traits. Yet, despite the substantial progress, notable gaps persist, often lurking in unexpected areas where our understanding had seemed comprehensive. One such gap pertains to binary traits, where even the fundamental task of sample size calculation has proven inadequate when relying on the most commonly used tool in genetic research, as shown in our recent study of the under-explored African sample within the UK Biobank data. Contrary to the analysis of continuous outcomes, disregarding the influence of covariates results in an underestimation of the required replication sample size, thereby exacerbating the perennial reproducibility crisis. The key statistical issue here is the non-collapsibility in logistic regression, a well-explored concept in the fields of social science and econometrics but largely overlooked in genetic studies. Unfortunately, we have recently demonstrated that failure to address non-collapsibility leads to biased effect size estimates when analyzing binary outcomes. Moreover, complex traits imply complex genetic models which include gene-environmental interactions. Yet, direct interaction modelling is often unviable due to the unavailability of data regarding the interacting environmental variables. When dealing with continuous traits, latent interactions introduce artificial heteroskedasticity in the outcome, a phenomenon that can be harnessed to facilitate latent or indirect interaction analyses. Subsequently, various methods, including our own, have been devised for continuous traits. However, many complex traits are not continuous, causing us to miss novel scientific discoveries if we do not develop a methodology for latent interaction analysis of binary traits. Finally, replicable and interpretable genetic studies require statistical methods that are not only powerful but also robust to model uncertainty and data heterogeneity. Given the ever-increasing complexities of large-scale genetic data being generated, method robustness will be a key consideration of this research program. Thus, my long-term goal is to develop advanced statistical genetics methods, robust against genetic model assumptions and missing data, with a focus on binary outcomes. Working towards this purpose, I will focus on two primary themes:    Theme 1: Address non-collapsibility in logistic regression in different study settings.    Theme 2: Develop latent or indirect interaction methods for binary traits when the standard direct interaction analysis is infeasible or undesirable. We will develop a catalogue of robust methods for different study settings, apply and test them on large-scale genetic data such as the UK Biobank data, and finally, disseminate the proposed methods as user-friendly software packages for researchers to use.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Ontario, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data

Tel que publié

Application id
RGPIN-2024-04665
Name nom
Sun, Lei
Department département
Statistical Sciences
Organization id
31
Institution établissement
University of Toronto
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
44000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1508
Committee name en
Mathematics and Statistics
Committee name fr
Mathématiques et statistique
Area of application code
1101
Area of application group en
Health, education and social services
Area of application group fr
Santé publique, éducation et services sociaux
Area of application en
Human health (including medically-related psychological research)
Area of application fr
Santé publique (y compris recherche psychologique liée à la médecine)
Research subject code
3004
Research subject group en
Statistics and probability
Research subject group fr
Statistique et probabilité
Research subject en
Applied statistics
Research subject fr
Statistique appliquée
Field of research list names en
Not available
Field of research list names fr
Non disponible
Socioeconomic objective list names en
Not available
Socioeconomic objective list names fr
Non disponible
Application title
Novel robust methods for complex binary traits
Keywords
statistical methodology; statistical genetics and genomics; applied statistics; robust association methods; binary outcomes; interaction; non-collapsibility; heterogeneity; meta-analysis; complex human traits
Application summary
In the last two decades, many advanced statistical methods have been proposed, developed and implemented for genetic studies of human complex traits. Yet, despite the substantial progress, notable gaps persist, often lurking in unexpected areas where our understanding had seemed comprehensive. One such gap pertains to binary traits, where even the fundamental task of sample size calculation has proven inadequate when relying on the most commonly used tool in genetic research, as shown in our recent study of the under-explored African sample within the UK Biobank data. Contrary to the analysis of continuous outcomes, disregarding the influence of covariates results in an underestimation of the required replication sample size, thereby exacerbating the perennial reproducibility crisis. The key statistical issue here is the non-collapsibility in logistic regression, a well-explored concept in the fields of social science and econometrics but largely overlooked in genetic studies. Unfortunately, we have recently demonstrated that failure to address non-collapsibility leads to biased effect size estimates when analyzing binary outcomes. Moreover, complex traits imply complex genetic models which include gene-environmental interactions. Yet, direct interaction modelling is often unviable due to the unavailability of data regarding the interacting environmental variables. When dealing with continuous traits, latent interactions introduce artificial heteroskedasticity in the outcome, a phenomenon that can be harnessed to facilitate latent or indirect interaction analyses. Subsequently, various methods, including our own, have been devised for continuous traits. However, many complex traits are not continuous, causing us to miss novel scientific discoveries if we do not develop a methodology for latent interaction analysis of binary traits. Finally, replicable and interpretable genetic studies require statistical methods that are not only powerful but also robust to model uncertainty and data heterogeneity. Given the ever-increasing complexities of large-scale genetic data being generated, method robustness will be a key consideration of this research program. Thus, my long-term goal is to develop advanced statistical genetics methods, robust against genetic model assumptions and missing data, with a focus on binary outcomes. Working towards this purpose, I will focus on two primary themes:    Theme 1: Address non-collapsibility in logistic regression in different study settings.    Theme 2: Develop latent or indirect interaction methods for binary traits when the standard direct interaction analysis is infeasible or undesirable. We will develop a catalogue of robust methods for different study settings, apply and test them on large-scale genetic data such as the UK Biobank data, and finally, disseminate the proposed methods as user-friendly software packages for researchers to use.