Loading...

A Library For Developing Optimization Algorithms In Metabolic Network Analysis

Ghadimi Deylami, Iman | 2023

55 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 56578 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Tefagh, Mojtaba
  7. Abstract:
  8. In systems biology, one of the most important biological systems that is analyzed and investigated is the metabolic network. A metabolic network is a complete set of metabolic and physical processes that determine the physiological and biochemical characteristics of a cell. These networks encompass metabolic chemical reactions, metabolic pathways, and regulatory interactions that govern these reactions. Therefore, metabolic networks at the genome scale are immensely large, making even efficient algorithms time-consuming for their analysis. To address this issue, reducing metabolic networks is crucial, as it significantly decreases the execution time of algorithms and enhances computational speed. Metabolic network reduction is a pertinent application of convex optimization in the analysis of metabolic networks. Our approach to solving this optimization problem is quantitative reductions. This concept is implemented through a constrained-based analysis method known as quantitative flux coupling analysis, and, by eliminating all possible reactions, ultimately lead to a reduced metabolic network. The first goal of this project is to design a Julia library (sparseQFCA.jl) to provide the necessary optimization algorithms for consistency checking and flux coupling analysis, and for the reduction of metabolic networks. Next, we aim to develop techniques for parallelizing these optimization algorithms and solving systems of linear equatios. In the final phase, we will run these algorithms on various systems biology datasets, including those from plants, animals, and bacteria. We will analyze the acceleration of the algorithms as well as the reduction in metabolic networks. In addition to providing more information content in the concepts of flux coupling and maximum reduction in metabolic networks, our method is significantly faster than previous methods and incomparable in terms of execution time. In the context of the consistency check algorithm without parallelization, it performs, on average, twice as quickly. Whereas, for other algorithms employing parallelization techniques, the speed enhancement is directly proportional to the count of threads. Additionally, this method has less numerical instability due to the implementation of optimization problems with Julia language
  9. Keywords:
  10. Metabolic Networks ; Convex Optimization ; Parallel Processing ; System Biology ; Optimization Algorithms

 Digital Object List

 Bookmark

No TOC