Timothy Frayling, Ph.D.
Professor of Human Genetics and Group Leader, University of Geneva (Geneva, Switzerland)
More about Timothy Frayling
Timothy Frayling is a molecular geneticist with over 20 years of experience
working on common and rare disease at the University of Exeter, UK. Since
2023, Timothy is leading a team at the Faculty of Medicine at the University
of Geneva, studying the genetics and genomics of common metabolic dis-
eases with a focus on type 2 diabetes, obesity and related conditions. He
earned his PhD in 1998 at the University of Exeter. His research teams have
become internationally recognised as world leaders in the genetics of com-
mon traits and conditions, contributing to the discovery of the first common
genetic variants linked to type 2 diabetes, fetal and childhood growth and
obesity.
Talk title
Using Genomic structural equation models, genome wide association study data
and simple clinical measures to provide an improved metric of obesity.
Authors
Tim Frayling, Daeeun Kim, Chi Zhao, Baiyu Qi, Emmaleigh Wilson, Gina Peloso,
Stavroula Kanoni, Cecilia Lindgren, Ruth Loos, Kristin Young, Kari North, Sonja Berndt,
Laura Raffield, Tuomas Kilpeläinen , Karen Mohlke, Anne Justice, Cassandra Spracklen,
Misa Graff on behalf of the GIANT and GLGC genome wide association study consortia
Abstract
People of the same weight, sex, age and ancestry can have very different
risks of obesity related disease, especially metabolic diseases such as type
2 diabetes and heart disease. These differences have led to proposals to im-
prove the definition of obesity so that it includes a better measure of disease
risk. Previous epidemiological studies have incorporated cut offs for waist
circumference and circulating factors such as glucose and triglyceride lev-
els to define “metabolic syndrome”. However, relying on thresholds for mul-
tiple risk factors wastes information when those risk factors are continuous
variables. Genomic structural equation models have become a popular and
robust method to incorporate information from multiple individual pheno-
types using summary data from large genome wide association studies
(GWAS). For example, a recent study used summary GWAS data from seven
metabolic traits and 4.9 million people to define a common latent trait for
metabolic syndrome. However, no such approaches have been designed to
explicitly capture the adverse effects of excess weight.
In this presentation I will discuss our Genomic SEM approach to provide a
better measure of the adverse effects of excess weight compared to BMI. I
will discuss how we used large scale summary GWAS data from four key
traits, BMI, waist hip ratio, circulating triglycerides and circulating High den-
sity lipoprotein levels, and Genomic SEM to generate a latent trait for meta-
bolically “unfavourable adiposity”. I will describe how we used sex and an-
cestry stratified data and will show how we tested the genetics of “unfavour-
able adiposity” in independent UK Biobank data. Finally I will describe how
we used downstream GWAS approaches to characterize the genetics of this
trait and compare it to BMI.
Joelle Mbatchou, Ph.D.
Senior Manager, Statistical Genetics, Regeneron Genetics Center (Tarrytown, New York, United States of America)
More about Joelle Mbatchou
Joelle Mbatchou, Ph.D., is a statistical geneticist at the Regeneron Genetics
Center, where she has worked since 2019. Her research focuses on devel-
oping statistical methods and computational tools for large-scale genetic
association analyses to better understand the impact of genetic variation on
human diseases. She has contributed to the development of tools like RE-
GENIE, which enable efficient modeling and analysis in large-scale biobanks.
Before joining Regeneron, Joelle earned her Ph.D. in Statistics from the Uni-
versity of Chicago where she developed statistical methods for genetic asso-
ciation analyses in structured samples. She also holds a dual B.Sc. in Biology
and Mathematical Sciences from DePaul University. Joelle is also involved in
teaching statistical genetics and genome-wide association studies, helping to
train the next generation of researchers in the field.
Talk title
Using Large Language Models for Rare Variant Association Testing in Large-Scale Biobanks
Abstract
The application of whole exome sequencing in studying of rare genetic vari-
ation has been well-established as a powerful and cost-effective strategy for
novel drug target discovery. The study of rare genetic variation, potentially
important in the development of complex diseases, has been increasingly
performed thanks to advances in sequencing technologies. Gene-based
tests have been developed to address the challenges with single variant
tests caused by the rarity of these variants and the need for large sample
sizes. These tests aggregate information across many variants and can
integrate external functional annotations to improve the power of rare var-
iant analysis. In recent years, large language models (LLMs) have been used
to predict the functional impact of genetic mutations, potentially enhancing
the power of rare variant association tests, and complementing functional
prediction approaches based on in-silico algorithms. We showcase the in-
tegration of functional scores leveraging LLMs for large-scale gene-based
association testing in the UK Biobank, highlighting their potential to improve
the detection of rare variant associations and advance our understanding of
complex genetic diseases.
Eleonora Porcu, Ph.D.
Genetics Senior Specialist, Nestlé Institute of Health Sciences (Lausanne, Switzerland)
More about Eleonora Porcu
Eleonora Porcu is a Genetics Senior Specialist at the Nestlé Institute of
Health Science in Lausanne. She holds a Master’s degree in Mathematics
from the University of Cagliari and a PhD in Biomedical Sciences from the
University of Sassari. Her research focuses on the development and appli-
cation of mathematical models to study the impact of the genetic makeup
on various aspects of nutrition and health.
Talk title
Omics and Mendelian Randomization: A Journey into Biological Mechanisms
Abstract
The increasing number of studies incorporating data across multiple biolog-
ical levels, such as genomics, transcriptomics, proteomics, and metabo-
lomics, raises critical biomedical questions regarding the systematic inte-
gration of these diverse datasets to uncover new biological mechanisms that
elucidate the processes of health and disease. Statistical causal frame-
works, particularly Mendelian randomization (MR), provide a robust founda-
tion for integrating these data and facilitating novel biological discoveries.
Mendelian randomization emerges as a valuable strategy for examining cau-
sality within complex biological and omics networks, offering insights that
can inform drug development and prioritize intervention targets for disease
prevention. Among the various omics fields, transcriptomics is the most ex-
tensively studied, with numerous investigations employing MR to identify
causal genes associated with complex traits and to differentiate between
mere correlations and true causal effects.
Recent advancements in multi-omics MR approaches have further inte-
grated additional omics layers, allowing researchers to explore the biolog-
ical pathways underlying gene-trait relationships. As these multi-omics
methodologies become increasingly prevalent, a more systematic approach
is essential to manage the growing complexity of data. Although combining
MR results with observational data can enhance the robustness of causal
inferences and provide a more comprehensive understanding of the relation-
ships between exposures and outcomes, biological validation remains a
critical step in confirming findings derived from these analyses.
While MR provides powerful tools for understanding causal relationships,
especially when randomized controlled trials are not feasible, the interpre-
tation of results must be approached with caution. Here I will highlight the
potential of multi-omics integration and MR in advancing our understanding
of health and disease, emphasizing the importance of rigorous validation
and careful interpretation in the pursuit of biological insights.
Marylyn Ritchie, Ph.D.
Chief Artificial Intelligence Officer, Director of a new Center for Al, Director of the Division of Computational Health Sciences and Al, College of Medicine, Medical University of South Carolina (Charleston, United States of America)
More about Marylyn Ritchie
Dr. Marylyn D. Ritchie is the Chief Artificial Intelligence Officer for the
MUSC Enterprise, Director of a new Center for Al, as well as Associate
Dean for Artificial Intelligence and Director of the Division of Computa-
Computational Health Sciences and Al in the College of Medicine at the
Medical University of South Carolina (MUSC). Dr. Ritchie is also the
SmartState Endowed Chair in Translational Biomedical Informatics.
Dr. Ritchie is an expert in translational bioinformatics, with a focus on de-
veloping, applying, and disseminating algorithms, methods, and tools
integrating electronic health records (EHR) with genomics. Dr. Ritchie has
over 20 years of experience in translational bioinformatics and has authored
over 500 publications. Dr. Ritchie was appointed as a Fellow of the American
College of Medical Informatics (ACMI) in 2020. Dr. Ritchie was elected as a
member of the National Academy of Medicine in 2021.
Talk title
Too Many Omics? There’s an App (Algorithm) for That
Abstract
The rapid growth of high-throughput technologies has produced an unprec-
edented wealth of multi-omics data, spanning genomics, transcriptomics,
proteomics, metabolomics, epigenomics, and beyond. While each individ-
ual layer offers valuable insights, the true promise of precision medicine and
complex trait discovery lies in their integration. Yet, bringing these heteroge-
neous, high-dimensional, and often noisy data sources together remains
one of the central challenges in mathematical genetics and computational
biology.
In this talk, I will explore a spectrum of machine learning approaches de-
signed for multi-omics integration, highlighting different machine learning
approaches with practical applications. I will begin with descriptions of the
different strategies for integration such as meta-dimensional and multi-
staged analyses. I will then move to specific applications where these ap-
proaches have been valuable in identifying important biological insights. Fi-
nally, I will discuss recent advances in deep learning, variational autoencod-
ers, and other latent factor models that offer flexible, nonlinear frameworks
for integration while grappling with issues of interpretability and generaliza-
bility. Throughout, I will emphasize the trade-offs between simplicity and
complexity, interpretability and predictive power, as well as supervised
versus unsupervised strategies. Case studies from population health and
disease genetics will illustrate how integrative analyses can uncover novel
biology that is invisible to single-omics approaches.
By the end, I hope to leave the audience with a practical “algorithmic toolkit”
for multi-omics integration, along with a sense of where the field is heading.
Whether the goal is risk prediction, biomarker discovery, or mechanistic in-
sight, machine learning offers increasingly powerful ways to make sense of
the overwhelming—and overlapping—layers of omics data.
Loading... 
