Principal metabolic flux mode analysis

Thumbnail Image

Files

master_Sourkatti_Heba_2018.pdf (2.52 MB)

URL

Journal Title

Journal ISSN

Volume Title

Perustieteiden korkeakoulu | Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.

Authors

Date

2018-11-07

Department

Major/Subject

Machine Learning and Data Mining

Mcode

SCI3044

Degree programme

Master’s Programme in Computer, Communication and Information Sciences

Language

en

Pages

49+7

Series

Abstract

In recent years, much progress has been achieved in the computational analysis of the metabolic networks, as a consequence of the rapid growth of the omics database. However, current literature analysis algorithms still lack good biological interpretability of the analysis results. Moreover, they can not be applied on a whole-genome level. This thesis assesses the potential of the Principal Metabolic Flux Mode Analysis (PMFA). The PMFA is a novel algorithm that was recently developed, which aims to improve the interpretability of Principal Component Analysis (PCA), through including a stoichiometric regularization to the PCA objective function. The PMFA can determine the flux modes that explain the highest variability in the network and it can also scale-up to a whole-genome level using the sparse version of PMFA. Furthermore, this thesis compares the PMFA to the recent approach Principal Elementary Mode Analysis (PEMA), which also tries to enhance the PCA interpretability. However, this approach is computationally heavy and thus fails to handle the large-scale networks (e.g., whole-genome). In order to further determine the feasibility of the PMFA approach for the analysis of metabolism, a Graph-regularized Matrix Factorization (GMF) was developed analogous to PMFA framework, similarly by adding the network stoichiometric matrix to a graph-structured matrix factorization framework. The results illustrate the potential of PMFA as a metabolic network analysis for identifying fluxes that explain maximum variation in the network and it can be used to analyze whole-genome level. In addition, the results showed that GMF method performed well in predicting active Elementary Modes (EMs) on simulated data but failed to work on large networks, while PEMA had the lowest performance among all methods. Based on the results, future work can be conducted to improve the GMF approach in terms of genome-scale analysis through including sparsity.

Description

Supervisor

Rousu, Juho

Thesis advisor

Bhadra, Sahely

Keywords

metabolic network analysis, PCA, stoichiometric modelling, matrix factorization

Other note

Citation

Permanent link to this item

https://urn.fi/URN:NBN:fi:aalto-201811135778

Collections

[dipl] Perustieteiden korkeakoulu / SCI