Document Type
Dissertation
Date of Award
5-31-2021
Degree Name
Doctor of Philosophy in Computer Engineering - (Ph.D.)
Department
Electrical and Computer Engineering
First Advisor
Ali Abdi
Second Advisor
Alexander Haimovich
Third Advisor
Reka Albert
Fourth Advisor
Joerg Kliewer
Fifth Advisor
Hongya Ge
Abstract
Developing molecular network analysis methods is important due to their applications on complex biological systems such as target discovery, development of drugs, discovering drug effects, and finding treatments for many complex diseases, e.g., cancer, autoimmune, and mental disorders. An example of analysis techniques is the fault diagnosis analysis, in which the purpose is to quantify how much vulnerable the entire network is to dysfunction of one or multiple molecules. Such analysis can be done after proper network models are implemented, trained, and tested against the experimental data. In this dissertation, a Boolean modeling framework is implemented and methods to train the models against data are presented on multiple networks. Furthermore, a mathematical framework for executing single and multi-fault vulnerability analysis of a given molecular network using the trained network models is provided. In addition, the worst possible signaling failures in molecular networks is examined by comparing the maximum vulnerability level, i.e., the highest probability of network failure, versus the number of faulty molecules to understand how the network functionality is affected in the presence of one or more dysfunctional molecules, for which an efficient algorithm is developed. Moreover, another algorithm is proposed that outputs the maximum number of time points needed for computing the vulnerability level of molecules in a Boolean domain. The methods are applied to the experimentally verified ERBB and T cell signaling networks. The results reveal that as the number of faulty molecules increases, the maximum vulnerability values do not necessarily increase, which means that a few faulty molecules can cause the most detrimental network damages and an increase in the number of faulty molecules does not deteriorate the network function. Such a group of molecules whose dysfunction causes the worst signaling failure may contribute to the development of the disorder and can suggest some therapeutic strategies.
Abnormality of a highly vulnerable molecule or a group of molecules results in incorrect network responses, which may cause the entire cell to make wrong decisions on the received signals and hence may initiate bigger events causing complex diseases. Therefore, characterization of decision-making in cells in response to received signals is of importance for understanding how cell fate is determined in the absence and presence of such abnormalities. Considering the cellular heterogeneity and dynamics of biochemical processes, the problem becomes multi-faceted and complex. This dissertation reveals a unified set of decision-theoretic, machine learning, and statistical signal processing methods and metrics to model the precision of signaling decisions in the presence of uncertainty, using single-cell data. This is done by presenting an optimal decision strategy minimizing the total decision error probability. Later, the framework is extended to incorporate the dynamics of biochemical processes and reactions in a cell, using multi-time point measurements and multidimensional outcome analysis and decision-making algorithms. Furthermore, the developed binary outcome analysis and decision-making approach is extended to more than two possible outcomes. As an example, and to show how the introduced methods can be used in practice, they are applied to single-cell data of PTEN, an important intracellular regulatory molecule in a p53 system, in wild-type and abnormal cells.
Recommended Citation
Ozen, Mustafa, "Modeling and analysis of intracellular signaling networks and cellular decisions" (2021). Dissertations. 1752.
https://digitalcommons.njit.edu/dissertations/1752
