Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

Identifying Patterns of Cancer Disease Mechanisms by Mining Alternative Representations of Genomic Alterations

Chen, Vicky (2017) Identifying Patterns of Cancer Disease Mechanisms by Mining Alternative Representations of Genomic Alterations. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

[img]
Preview
PDF
Download (4MB) | Preview

Abstract

Cancer is a complex disease driven by somatic genomic alterations (SGAs) that perturb signaling pathways and consequently cellular function. Identifying combinatorial patterns of pathway perturbations would provide insights into common disease mechanisms shared among tumors, which is important for guiding treatment and predicting outcome. However, identifying perturbed pathways is challenging, because different tumors can have the same perturbed pathways that are perturbed by different SGAs.
We started off by designing a novel semantic representation that captures the functional similarity of distinct SGAs perturbing a common pathway in different tumors. This representation was used alongside the nested hierarchical Dirichlet process topic model in order to identify combinatorial patterns in altered signaling pathways. We found that the topic model was able to capture the functional relationships between topics. It was also able to identify cancer subtypes composed of tumors from different tissues of origin that exhibit different survival rates.
These results led us to investigate the performance of the methodology on pan-cancer data, as well as in conjunction with cancer driver data. The results revealed that the framework was still able to identify clinically relevant features in pan-cancer. However, the addition of driver data decreased the noise in the data and improved the separation of tumors in the clustering results. This provided support for including the use of driver data in our methodology.
In order to have gene representations independent of literature, we developed a biological representation that could identify functionally related genes. Its performance when used alongside topic modeling was tested. We found that the topic association patterns separated tumors by their tissue of origin. But, analyzing some of the cancer types on an individual basis still led to significant differences in survival.
Our studies show the potential for using alternative representations in conjunction with topic modeling to investigate complex genomic diseases. With further research and refinement of this methodology, it has the potential to capture the relationship between pathways involved in cancer. This would contribute to a better understanding of cancer disease mechanisms and treatment.


Share

Citation/Export:
Social Networking:
Share |

Details

Item Type: University of Pittsburgh ETD
Status: Unpublished
Creators/Authors:
CreatorsEmailPitt UsernameORCID
Chen, Vickyvic14@pitt.eduvic140000-0003-4595-154X
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairLu, Xinghuaxinghua@pitt.edu
Thesis AdvisorLu, Xinghuaxinghua@pitt.edu
Committee MemberJacobson, Rebeccarebeccaj@pitt.edu
Committee MemberHochheiser, Harryharryh@pitt.edu
Committee MemberPaisley, Johnjpaisley@columbia.edu
Date: 18 January 2017
Date Type: Publication
Defense Date: 13 September 2016
Approval Date: 18 January 2017
Submission Date: 17 November 2016
Access Restriction: No restriction; Release the ETD for access worldwide immediately.
Number of Pages: 141
Institution: University of Pittsburgh
Schools and Programs: School of Medicine > Biomedical Informatics
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: cancer, topic modeling, semantic representation, biological representation, cancer genomics, disease mechanisms
Date Deposited: 18 Jan 2017 15:45
Last Modified: 19 Jan 2017 06:15
URI: http://d-scholarship.pitt.edu/id/eprint/30319

Metrics

Monthly Views for the past 3 years

Plum Analytics


Actions (login required)

View Item View Item