Context-Specific Reconstruction and Gap-Filling of Metabolic Networks by Sparse Reconciliation of Data Inconsistencies

Fathi, Ali | 2022

104 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55304 (02)
  4. University: Sharif University of Technology
  5. Department: Mathematical Sciences
  6. Advisor(s): Tefagh, Mojtaba
  7. Abstract:
  8. With the increasingly collected biological data, appropriate usage of this data is of great importance for understanding and predicting biological systems and has been the aim of experiments and data collections. A famous category of biological data is known as “omics” which refers to transcriptomics, proteomics, metabolomics, and fluxomics, from different cells or tissues in various media and conditions. This set of data is regularly used for tasks such as studying cells and organisms, understanding cell states, cancer prediction, etc. and is of great importance in Systems Biology.In this thesis, we concentrate on studying cells or organisms using such data, where during that process, we reconstruct the organism's metabolic network with the appropriate use of the data. By metabolic network, we mean the set of all existing and active reactions inside the desired cell with which metabolites are converted to each other.The method we concentrate on here in this study to do such is using “sparse optimization”. In this method, we try to find a minimal set of reactions that are consistent with the maximum number of available data of the cell or organism of interest. By sparsity, we mean that the model should utilize as few as possible reactions from the list of possible biological reactions.For biological reasons, this minimality of the metabolic network is essential because whenever a cell or an organism does not need a reaction, it tries to get rid of that reaction or reduce the expression of its related genes, as maintaining such a reaction overburdens the organism with a load. Also, this minimality helps the reconstructed network to satisfy as much as possible no-growth conditions data, which is used as the primary validation for the accuracy of our method. Besides, we emphasize another strength of our method in this thesis, which is simultaneously utilizing a wide variety of data, a strength commonly set aside in related works. Only one type of data has usually been used for metabolic reconstruction in those works.As the result of using the proposed algorithm to reconstruct the metabolic network of a famous bacteria called “Bacillus subtilis”, our method successfully identified 86 percent of reactions in the metabolic network of the bacteria by utilizing phenotypic data, which is an 11 percent improvement compared to the exclusive utilization of the genome sequence. An important note here is our method is fully automatic, and achieving greater than 11 percent improvement has been possible only by paying extensive cost and time of so many experts. All implementations and also codes for automatic preprocessing done on biological data are accessible on the page https://github.com/Alef125/SparseReconstruction.

  9. Keywords:
  10. Sparse Optimization ; Metabolic Networks ; Phenotypic Data ; Metabolic Network Reconstruction ; Network Reconstruction ; Metabolic Network Gap-Filling

 Digital Object List