Steiner Trees and Multilayer Networks for Functional Analysis and Integration of High-throughput Omics Data in Cancer Systems Biology

Decanato - Facoltà di scienze informatiche

Data: 4 Ottobre 2018 / 10:30 - 12:00

USI Lugano Campus, room CC-351, Main building (Via G. Buffi 13)

You are cordially invited to attend the PhD Dissertation Defense of Murodzhon AKHMEDOV on Thursday, October 4th, 2018 at 10h30 in room 351 (main building)



The initiation and progression of complex diseases such as cancer is caused by the accumulation of multiple aberrations in different genes. Systems biology is an emerging research field to understand the biological functions of genes using multiple high-throughput omics data types. However, integrative analysis of heterogeneous omics data types is currently a challenging task in systems biology. Omics data integration is essential to obtain a comprehensive overview of otherwise fragmented information and to better understand dysregulated biological pathways leading to a particular condition. The primary goal of this thesis is the development and application of computational approaches for functional analysis and integration of multiple omics data based on graph-related methods.

The Steiner tree problem is well known in graph theory and combinatorial optimization. Here, we concentrate on its application in biology to perform functional analysis and to identify functional components. We develop heuristic and matheuristic solution strategies for solving large biological network instances. These methods are rigorously tested on benchmark instances and their computational performances are compared to the methods from the literature. Biological applications of the methods are also demonstrated, such as identifying functional units in a given network interactome and selecting signature genes that further classify cancer patients into subtypes.

Furthermore, we develop an R-package based on our Steiner tree method. The package performs a fast and user-friendly network analysis of high-throughput data by mapping the data into biological networks.

We also propose a novel multilayer network framework for the integration and analysis of multi-omics data of heterogeneous types. The multilayer framework is composed of omics layers, functional layers and phenotype layer, where each omics layer represents a certain measured data type. Functional layers may represent genesets, pathways or biological concepts to facilitate functional interpretation of the data, while phenotype layer harbours the clinical data of patients. The framework then calculates the highest coefficient paths in a multilayer network from each omics feature to the phenotype by computing an integrated score along the paths. These paths may indicate the most plausible signalling cascade caused by perturbed omics features leading to a particular phenotype response. With example applications, we demonstrate the potential power of our method in functional analysis and biomarker discovery in personalized medicine using multi-omics data.


Dissertation Committee:

  • Prof. Roberto Montemanni, Istituto Dalle Molle di Studi sull’Intelligenza Artificiale, Switzerland (Research Advisor)
  • Prof. Luca Maria Gambardella, Università della Svizzera Italiana, Switzerland (Co-Advisor)
  • Dr. Ivo Kwee, Istituto Dalle Molle di Studi sull’Intelligenza Artificiale, Switzerland (Co-Advisor)
  • Prof. Laura Pozzi, Università della Svizzera italiana, Switzerland (Internal Member)
  • Prof. Olaf Schenk, Università della Svizzera italiana, Switzerland (Internal Member)
  • Prof. Mauro Delorenzi, Swiss Institute Bioinformatics, Switzerland (External Member)
  • Prof. Vittorio Maniezzo, University of Bologna, Italy (External Member)