Integrative Modeling To Predict Dynamics-Based Therapeutic Strategies: A Case Study With ALS
Author(s)
Bi, Sarah Dongyun
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
Amyotrophic lateral sclerosis (ALS) is a fatal disease characterized by motoneuron degeneration in the brainstem, spinal cord, and cortex, which can lead to muscle paralysis, dysphagia, respiratory distress, and ultimately death. Despite extensive studies of ALS, no cures have been discovered. Furthermore, only two ALS treatments have been approved by the United States Food and Drug Administration (FDA). Thus, there exists a dire need to explore better treatments for ALS.
Superoxide dismutase 1 glycine 93 to alanine (SOD1-G93A) transgenic mice exhibit mutations associated with ALS. However, it is difficult and resource-intensive to perform high throughput combination treatment testing for ALS in vivo. The primary study objective was to model targets for ALS combination therapy in silico using stability as the criterion for a successful treatment target. Computational models of wild type (WT) and SOD1-G93A (ALS) mouse physiological stability were developed using dynamic meta-analysis (DMA), mathematical optimization, and unsupervised machine learning, based on prominent regulatory mechanisms in ALS pathophysiology. Combination treatment targets based on these regulatory mechanisms were tested to simulate ALS progression and treatment. The identified treatment targets and stability patterns provide critical insight for experimental development and testing of translational ALS treatments and etiological hypotheses. This study introduces an innovative method to create a mathematical model of multiscalar, multifactorial disease to assess ALS combination treatment performance.
Literature based discovery (LBD) is a text mining technique that extracts and integrates knowledge from existing literature and seeks to discover new relationships by examining patterns in a knowledge graph containing text-mined relationships from millions of PubMed articles spanning all biomedical domains. Combination treatment targets identified from the SOD1-G93A computational mouse model were used as starting points for LBD. The goal of LBD was to recommend repurposed drugs for ALS that match the mechanistic intent of the identified highly ranked dynamics-based combination therapies. In short, LBD provides connections for translational actionable insight for the treatment of ALS.
Finally, the mouse data collected for this study was manually curated from biomedical text. However, manual curation of biomedical text is an expensive and time-consuming task. Natural language processing (NLP) was used to create an automated tool that aids the curation of biomedical articles. This tool allows for faster data extraction from biomedical literature and can lower the burden of data collection for similar computational models in the future.
Sponsor
Date
2022-04-27
Extent
Resource Type
Text
Resource Subtype
Thesis