Events Calendar

14 Dec
Biostatistics Seminar - Rebecca Deek, PhD Candidate, University of Pennsylvania Perelman School of Medicine
Event Type

Lectures, Symposia, Etc.

Target Audience

Faculty, Graduate Students, Postdocs

University Unit
Department of Biostatistics
Subscribe
Google Calendar iCal Outlook

Biostatistics Seminar - Rebecca Deek, PhD Candidate, University of Pennsylvania Perelman School of Medicine

This is a past event.

Mixture margin copula models for inferring covariation networks with applications in microbiome and single cell genomics

 

Abstract:

Covariation networks among bacterial taxa in microbial communities provide important insights into the connectivity and structure of microbial ecosystems. Similarly, co-expression networks among genes, at the cellular level, help to identify the gene-gene relationships that shape cell identities. High-throughput sequencing technologies enable researchers to measure microbial composition and single cell gene expression. The resulting data are often characterized by excessive zeros, sequencing depth constraints, and high dimensionality. Existing methods using simple correlations and/or log-based transformations cannot capture the covariations observed in real data sets. I will present two copula models with mixture margins to estimate the covariation between a pair of zero-inflated variables. The proposed models allow for easy adjustment of observed confounders and can be applied to longitudinal microbiome data to construct temporally conserved covariation networks. I will also present inference methods using two-stage maximum likelihood and the Monte Carlo EM algorithm. These methods will be demonstrated using data from the American Gut and the longitudinal DIABIMMUNE projects. Finally, I will introduce some of my emerging and future research topics.

Wednesday, December 14 at 3:30 p.m. to 4:30 p.m.

1155 Public Health

Biostatistics Seminar - Rebecca Deek, PhD Candidate, University of Pennsylvania Perelman School of Medicine

Mixture margin copula models for inferring covariation networks with applications in microbiome and single cell genomics

 

Abstract:

Covariation networks among bacterial taxa in microbial communities provide important insights into the connectivity and structure of microbial ecosystems. Similarly, co-expression networks among genes, at the cellular level, help to identify the gene-gene relationships that shape cell identities. High-throughput sequencing technologies enable researchers to measure microbial composition and single cell gene expression. The resulting data are often characterized by excessive zeros, sequencing depth constraints, and high dimensionality. Existing methods using simple correlations and/or log-based transformations cannot capture the covariations observed in real data sets. I will present two copula models with mixture margins to estimate the covariation between a pair of zero-inflated variables. The proposed models allow for easy adjustment of observed confounders and can be applied to longitudinal microbiome data to construct temporally conserved covariation networks. I will also present inference methods using two-stage maximum likelihood and the Monte Carlo EM algorithm. These methods will be demonstrated using data from the American Gut and the longitudinal DIABIMMUNE projects. Finally, I will introduce some of my emerging and future research topics.

Wednesday, December 14 at 3:30 p.m. to 4:30 p.m.

1155 Public Health

Target Audience

Faculty, Graduate Students, Postdocs

University Unit
Department of Biostatistics

Powered by the Localist Community Events Calendar ©