thesisAbstract.txt

Hepatocellular carcinoma (HCC) is the most common form of liver cancer, which is the second most prevalent cause of cancer mortalities worldwide. 
The current standard of care for non-resectable HCC is to combine immunotherapy with targeted therapy, yet the median overall survival is only 19 months. 
A reasonable question is whether survival could be increased if drugs developed for other purposes were repositioned to support HCC therapy. 
The Broad Institute�s Connectivity Map (CMap) provides an approach to drug repositioning, essentially translating between diseases, drug compounds, and cellular components. 
Given a gene signature of a disease, CMap scores compounds within its databases to identify drugs that would most effectively oppose it.
Most CMap studies are based on disease signatures derived from either bulk tissue or cell lines. 
Bulk tissue studies co-mingle signals from various cell types, obfuscating the underlying complexity and contributions of the tumor microenvironment (TME), which contains a wide variety of stromal and immune cells. 
Cell line studies present difficulty in modeling the TME�s full range of signals over time. 
Drug resistance often challenges therapy, and has been attributed to the dynamic complexity of the TME. 
CMap could yield more insightful results if applied in the context of single-cell RNA-sequencing (scRNA-seq). 
Single-cell analysis allows characterization of cell identity and function by attributing details of both malignant cells and the TME to their specific lineages. 
The impact of this attribution is to isolate the TME�s tumor-supporting and tumor-suppressing mechanisms from the malignancy itself.
Using previously published scRNA-seq data from 13 HCC patients and a normal tissue donor, the current study identifies 12 key cell types, recognizing both inter- and intra-tumor heterogeneity. 
When surgical tumor sections are contrasted with sections of healthier tissue, 6428 unique genes are differentially expressed. 
Requirements of CMap online processing, together with inference of the desirable direction of drug action, reduce these 6428 targets to a core set of 1529 for drug screening. 
This screening results in a set of 425 compounds for which the drug�s disease-suppressing mechanisms of one cell type appear to be unopposed by disease-promoting mechanisms of others. 
Using both CMap and scRNA-seq, the current study demonstrates a proof-of-concept for drug discovery tools based on in silico analysis, identifying drugs as candidates for further study.