Doctor of Philosophy in Bioinformatics

in the School of Biological Sciences

Xingyu Yang

Defends his thesis:
Computational approaches to intuitively analyze and visualize big data in biological research

Tuesday, November 6, 2018

2:00pm
Materials Science and Engineering Building 3201a

Thesis Advisor:

Dr. Peng Qiu
Department of Biomedical Engineering
Georgia Institute of Technology and Emory University

Committee Members:

Dr. Soojin Yi
School of Biological Sciences
Georgia Institute of Technology

Dr. Gregory Gibson

School of Biological Sciences
Georgia Institute of Technology

Dr. David Archer

Department of Pediatrics

Emory University School of Medicine

Dr. Ignacio Sanz
Department of Medicine

Emory University

Abstract:

To quantitatively understand the cell behavior in molecular level, scientists have developed technologies including high throughput sequencing and flow cytometry. High throughput sequencing can obtain the entire genome sequence and measure expression of large number of genes. Flow cytometry can measure multiple parameters of large number of cells. Both technologies generate large amount of data in high dimension. Therefore, efficient methods to analyze and interpret the data become in demand. In my thesis, I focus on developing computational methods that deliver intuitive and interpretable visualization of biological data. The first chapter describes a software named Cluster-to-Gate (C2G) that can visualize existing clustering results of flow/mass cytometry data in the format of 2D gating hierarchy. Though C2G presents a way to visualize and interpret clustering results, the visualization is still data-driven and no human-knowledge is incorporated. To overcome the limitation of C2G, the second chapter describes a framework that can learn gating approach from existing publications to build a knowledge-graph. This knowledge-graph can automatically suggest order of marker usage and gating hierarchy for new data set, which can be used to gate cell populations. The obtained cell populations are immediately matched to known cell types in the knowledge-graph, which makes them interpretable. The third chapter describe a novel algorithm (GLaMST) to reconstruct lineage tree of B cell receptor gene from high throughput sequencing data. This algorithm outperforms state-of-art in both accuracy and speed.