Some Clustering Approaches for High-dimensional Data

This dissertation is inspired by previous clustering research in recurrent event data and genomic expression data, with the goal of incorporating new features to extend existing methods and derive meaningful conclusions. In the first framework, we proposed a nested clustering model for recurrent event data to capture the heterogeneity structure in both patients and multi-type events while adjusting for confounding covariates. We also develop a computationally efficient algorithm for fast model estimation based on variational inference. In our second framework, We achieve temporal gene expression profile clustering based on treatment control differences and account for the covariance structure between repeated measurements of the same gene over time and between biological replicates. An Rcpp implementation of these methods is provided for efficient computation. For each framework, simulation studies demonstrate that our methods accurately recover clustering structure and outperform existing alternatives. In real-world applications, we apply the recurrent-event clustering approach to ICU electronic health-record dataset and the temporal profile clustering approach to a published microarray transcriptomic dataset—each yielding interpretable and meaningful results.

Author dc:creator

• Liyong Cui (23292022)

• Biology
• Biostatistics

Statement dc:rights

• In Copyright
• Open Access after 2028-01-01

DOI dc:identifier

https://doi.org/10.25417/uic.31451692.v1

OAI identifier oai:identifier

oai:figshare.com:article/31451692