Elucidating complex biological interactions using computational techniques

Fan, Zhenjiang (2023) Elucidating complex biological interactions using computational techniques. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

Preview	PDF (Main Dissertation Content) Primary Text Download (3MB) | Preview
	Microsoft Excel (Supplemental Table 1) Supplemental Material Download (551kB)
	Microsoft Excel (Supplemental Table 2) Supplemental Material Download (93kB)

Abstract

Studying complex biological systems faces numerous technical challenges due to their intricate nature and the multitude of interacting factors involved while analyzing related datasets. These challenges include the data diversity in biomedical datasets, nonlinearity behaviors among variable interactions, contextual causal factors, subtype heterogeneity, and causal mechanism complexity. To address these challenges, we must build specified computational models to tackle certain problems. Two of the most widely used computational tools are machine learning (ML) and deep learning (DL). Despite their tremendous potential, integrating ML and ML into biological research is not a trivial task. In our first project where we aim to understand dynamics among complex biological networks, such as subtype biological networks for a disease, we utilized a network similarity measuring method based on normalized Laplacian matrix eigenvalue distribution to systematically identify a comparable estrogen receptor negative (ER-) normal ceRNA network comparable to estrogen receptor positive (ER+) normal reference ceRNA network. We exploited various network analysis techniques to study dynamics among constructed subtypes of breast cancer. Our systematically analyzing disease subtype network using these network analysis techniques provides a meaningful research direction. For our second project where we determine to address the nonlinearity behavior and identify complex causal mechanisms in complex biomedical data, we developed a causal inference
v
method that learns both linear and nonlinear causal relations and estimates the effect size using a deep-neural network approach coupled with the knockoff framework. By using both simulation data and multiple real world biomedical datasets, we demonstrated that our proposed method outperforms existing methods in identifying true and known causal relations. The identified nonlinear causal relations and estimating their effect size can help understand the complex disease pathobiology, which is not possible using other methods. In our third project where we aim to address the data diversity, nonlinearity behavior, contextual causal factor problems in single-cell sequencing datasets, we created a DL model to identify condition-specific cell subtypes when we have multiple types of information. In comparison with existing clustering algorithms, our proposed clustering method outperforms them in terms of various evaluation matrices using both simulation data and real-world single-cell sequencing data.

---

Share

Citation/Export:	Select format... Citation - Text Citation - HTML Endnote BibTex Dublin Core OpenURL MARC (ISO 2709) METS MODS EP3 XML Reference Manager Refer
Social Networking:	Share |

---

Details

Item Type:	University of Pittsburgh ETD
Status:	Unpublished
Creators/Authors:	Creators Email Pitt Username ORCID Fan, Zhenjiang zhf16@pitt.edu zhf16 0000-0002-5889-5340
ETD Committee:	Title Member Email Address Pitt Username ORCID Committee CoChair Park, Hyun Jung hyp15 Committee CoChair Lee, Stephen stephen.lee Committee Member Huang, Heng heng.huang Committee Member Kovashka, Adriana aik85 Committee Member Tang, Xulong xulongtang
Date:	6 September 2023
Date Type:	Publication
Defense Date:	31 July 2023
Approval Date:	6 September 2023
Submission Date:	6 August 2023
Access Restriction:	No restriction; Release the ETD for access worldwide immediately.
Number of Pages:	158
Institution:	University of Pittsburgh
Schools and Programs:	Dietrich School of Arts and Sciences > Computer Science
Degree:	PhD - Doctor of Philosophy
Thesis Type:	Doctoral Dissertation
Refereed:	Yes
Uncontrolled Keywords:	Computational biology, network comparison, machine learning, deep neural network, causal inference, clustering, single cell
Date Deposited:	06 Sep 2023 15:36
Last Modified:	06 Sep 2023 15:36
URI:	http://d-scholarship.pitt.edu/id/eprint/45308

Available Versions of this Item

• Elucidating complex biological interactions using computational techniques. (deposited 06 Sep 2023 15:36) [Currently Displayed]

---

Metrics

Monthly Views for the past 3 years

Plum Analytics

---

Actions (login required)

View Item