Publication Date




Embargo Period


Degree Type


Degree Name

Doctor of Philosophy (PHD)


Human Genetics and Genomics (Medicine)

Date of Defense


First Committee Member

Nagi G. Ayad

Second Committee Member

Derek M. Dykxhoorn

Third Committee Member

Stephan C. Schürer

Fourth Committee Member

Jennifer L. Clarke

Fifth Committee Member

Claes Wahlestedt

Sixth Committee Member

Sunil J. Rao


Glioblastoma is the most common malignant primary adult brain tumor with a standard of care consisting of maximal surgical resection followed by radiotherapy and adjuvant temozolomide (TMZ) chemotherapy. However, despite medical advances in the field, recurrence is almost universal. As with most cancers, heterogeneity and adoptive reprogramming upon therapy, which often leads to resistance, represent huge barriers to clinical care. Novel targeted therapies, which are the foundation of precision medicine, are therefore urgently required. During the last decade, several large research consortia have generated unprecedented amounts of data to characterize and model complex biological systems and disease, in particular, cancer. Consequently, ‘big data’ approaches are now emerging in biomedical research. However, significant data science challenges still exist to fully leverage these resources. In this work, several informatics and computational solutions were developed and implemented to integrate, analyze and model data generated in the Library of Integrated Network-based Cellular Signatures (LINCS) Consortium. This work formed the foundation for the development of a general approach to prioritize and rank glioblastoma combination therapies. In addition, the data science work contributed to several tools enabling the community in data-driven research projects. Towards the development of glioblastoma therapies, transcriptional data from patient samples were used to generate personalized gene co-expression networks. The networks were further characterized and validated using available protein-protein interaction and biochemical data. Following these initial results to identify prospective personalized therapies, the glioblastoma data was then integrated with the LINCS L1000 transcriptional perturbation response library. Analyzing the L1000 data across many cell lines and small molecules, LINCS compounds could be clustered into distinct pharmacological drug classes based on the transcriptional changes they induce and the pathways they modulate. The integrative data analytics approaches were facilitated by the rich metadata and curated LINCS compound target annotations. These results led to the hypothesis that compounds affecting orthogonal transcriptional pathways, or distinct co-expression networks, would synergize in reducing proliferation of glioblastoma cells. The top compound combination was selected and tested in in vitro and in vivo assays validating this approach. In summary, foundational data science work and novel computational biology algorithms enabled the integration and modeling of data from several distinct sources providing a novel platform for identifying therapeutic combinations in glioblastoma and other cancers.


Cancer; Big Data; Combination Therapy; LINCS; TCGA; Personalized Therapies