Investigating determinants of tumor escape and developing computational tools for protein-protein interface analysis
| Autor: | Ghani, Usman |
|---|---|
| Rok vydání: | 2022 |
| Předmět: | |
| Druh dokumentu: | Thesis/Dissertation |
| Popis: | Tumor formation and progression are controlled by both extrinsic changes in the tumor microenvironment and intrinsic changes in individual tumor cells. For its part, the extracellular matrix (ECM) in the microenvironment can exert significant control through not only its composition but also through its biophysical properties. Though a lot is known about the effect of ECM on invasion, the same is not true about its effect on intravasation. In the first part of this work, we examined the role of biophysical properties of the ECM on tumor intravasation. Specifically, we developed a 3D in vitro model for the step after invasion in the metastatic cascade where tumor cells escape into the lumen of a vessel which, in our model, is simulated using an empty cavity. We looked at how ECM density, stiffness, and permeability affect tumor behavior and show that lowering density encourages tumor escape by increasing permeability of the matrix. With regards to intrinsic changes in the tumor cell, upsetting of protein-protein interactions (PPI) has been implicated in cancer progression. Modulation of PPIs thus provides a potential strategy for treatment. Accordingly, there has been much interest in finding and targeting binding sites at the interface of interacting proteins. In the second part of this work, we used a computational hotspot mapping tool to predict potential binding sites and druggability of proteins participating in PPIs. We found that hotspot mapping was able to predict inhibitor binding sites for a majority of our proteins of interest. The study of PPIs requires the structure of the protein-protein complex to be known, so we also developed and assessed computational tools for the prediction of protein-protein complex structures for cases where the structure is not available. We show that introducing physics based information to machine learning tools improves prediction quality, especially in cases where coevolutionary information is not available. 2024-08-26T00:00:00Z |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|