Browsing by Subject "computational Biology, Bioinformatics"

Sort by: Order: Results:

Now showing items 1-1 of 1
  • He, Liye (Helsingin yliopisto, 2020)
    Cancer is a dynamic disease driven by complex molecular and environmental interactions. Therefore, the traditional “one gene, one drug, one disease” strategy is insufficient to treat most cancer patients. Drug combination therapy targeting various molecular mechanisms has become increasingly popular in treating cancer and other complex diseases. In comparison to monotherapy, combination therapy has the following advantages: possibility to reduce the dose of each single drug to minimize toxic side effects; achieving at least additive, multi-targeting effects, or even “greater-than-additive” effects, so called synergy; and reducing the likelihood of treatment resistance. However, even with the advanced high-throughput technologies currently used in drug combination screening, it remains infeasible to test systematically all the possible drug combinations across different cancer types, as the number of combination experiments grows exponentially. In addition, it remains difficult to understand the synergy mechanisms of many drug combinations, which poses further challenges for their clinical approval. Therefore, there is a timely need for computational tools that can help in identifying synergistic drug combinations for each individual patient, revealing the mechanisms of action of the drug combinations in the specific cellular context, as well as discovering potential predictive biomarkers for the synergistic responses in a systematic and reproducible way. In this thesis, I implemented a systematic computational framework for identification and validation of synergistic drug combinations for each individual patient. Firstly, I developed machine learning models to predict drug combination effects by utilizing drug-target interaction networks, drug sensitivity profiles as well as genomic profiles of each patient. The models further enable one to identify both synergistic and cancer-selective drug combinations specific for each patient, therefore avoiding broadly toxic combinations. Secondly, to experimentally validate the predicted synergistic drug combinations, I developed software tools to help designing multi-dose drug combination experiments, to facilitate the semi-automated drug screening process, as well as to analyze the high-throughput drug combination screening data. Thirdly, I demonstrated the potential of constructing patient-specific cancer vulnerability networks to investigate the mechanisms of action of the personalized drug combinations, which is of importance to accelerate anticancer therapy discovery for precision oncology. All of the computational tools developed in this thesis are distributed as open-source tools, making it possible for others to reproduce the results and apply to their data, or to extend the tools to new applications, as well as to integrate the tools as part of in-house drug combination analysis pipelines.