Masters candidate in Biostatistics, Yan Li, will present:
Plan B Adviser: Baolin Wu
Abstract: The emerging technique, Single cell RNA sequencing (scRNA-seq), has brought promise in identifying cell-to-cell variation, which makes a full understanding of the molecular mechanism possible in many complex and evolving physiological as well as pathological processes, like embryonic development and cancer. Here, we propose a novel model based dimensionality reduction approach specifically designed for droplet-based scRNA-seq data, named as NB-WSVD, which employs negative binomial model and can adjust for the variation in detected genes per cell. When comparing our proposed method to the widely used principle component analysis (PCA) approach in both simulated and real data analysis, we found that PCA has better clustering effect in less sparse (percent of zero entries around 80%) scRNA-seq data, but NB-WSVD tends to perform better than PCA in moderately or extremely sparse scRNA-seq data (percent of zero entries around 90%).
Refreshments will be served prior to the presentation.