Deconvoluting complex disease mechanisms via machine learning methods for targeted therapeutics
Year: 2024
Type: dissertation
Abstract: Polypharmacology provides an alternative to investigate mechanisms of complex diseases like cancer. Various biological processes are involved in the development of cancer, forming multilevel heterogenous networks. Unveiling the complex interplays between components in such networks may lead to better understanding of disease progression and target identification. Better therapeutics can then be discovered by targeting selectively the desired components in the disease network, for example single drug therapies with designed polypharmacological profiles. Furthermore, combination therapies are promising as well due to their synergistic effects with reduced dosages. However, both target identification and drug discovery are limited by extensive laboratory experiments required for finding meaningful leads. Computational tools utilizing accumulated biological data provide powerful tools to narrow down the search space and thus accelerate such discovery process. In this thesis, various computational tools are developed to aid target identification and drug discovery. First, we implemented a computational framework to link drug-cell responses and drug-target interactions, for finding cancer cell-specific key targets and to better understand the potential mechanisms of cancer progression. We designed a statistically stable target addiction score to evaluate targets involved in the cancer cell responses to drugs. Second, we designed a novel target-specific selectivity metric, to better evaluate the polypharmacological profiles of drugs and aid multi-target drug design and repurposing. Third, we built state-of-art machine learning methods to predict potential synergistic drug combinations. Predictive models for drug combination responses were built based on polynomial regression via latent tensor reconstruction, which exhibited high performance in various application scenarios.
Source: Työväentutkimus Vuosikirja
Author: Wang, Tianduanyi
Language: English
Subfield: Drug Discovery
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7197388604