Loading...
Probing the Relation Between Cell Function and Chromatin Interactions Using Community Detection Approaches
Ghavami Modegh, Rassa | 2018
709
Viewed
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 51316 (19)
- University: Sharif University of Technology
- Department: Computer Engineering
- Advisor(s): Rabiee, Hamid Reza; Jafari Siavoshani, Mahdi
- Abstract:
- Hi-C is a new technology invented to record the amount of interaction between all chromatin fragments. Chromatin interactions are manifestations of chromatin's 3D structure. It has been proved that different 3D structures of different cell types, despite having the same DNA, causes in different functionalities of cells. Therefore knowing chromatin structure, the factors affecting it and the way it affects cells' behaviors such as gene co-expression, sheds more light on the knowledge about cells and the methods that can change their fates. Finding the causes of diseases, designing more efficient drugs and progressing the conversion of different cell types into each other used for amending damaged tissues are among the utilization. Hi-C is a complex, multilevel technology and the results of each level are affected by the biases related to that level. Furthermore, in addition to the real interactions, so many unreal interactions caused by random collisions of fragments during their Brownian motion, are also recorded in the technology. Sound analyzing and conclusion require the elimination of biases and unreal interactions. Due to the problems in the theory of the current methods, a new background correction and normalization model, devoid of the problems of previous methods, has been proposed. Operation of the proposed model has been compared with two of the newest background correction models and its superiority in performance has been proved. Next, the output of Hi-C in 11 samples, 6 from one cell type and each of the 5 others from a different cell type, have been modeled as a complex network and community detection approaches have been used to find the groups of related chromatin fragments in each cell type. With comparing the percentage of shared communities of the networks related to one cell type and the ones related to different cell types, considering the somehow low percentage of shared significant interactions between some of the samples, it has been proved that community detection approaches act more robustly in detecting important interactions and despite different protocols and lab conditions of Hi-C samples, radical group interactions remain persistent in samples. With comparing the percentage of shared communities related to replicated Hi-C samples with other samples, high variability of Hi-C samples and insufficiency of one sample in knowing one cell type thoroughly, have been proved. Next, with the goal of aggregating different genomic data on the output of Hi-C, for enhancing detected communities and inclining them toward detecting the relation between genes, a multi-layer network has been proposed that models the cooperation of genes, transcription of each gene and the effects of genes on each other. The detection of 138 communities with gene groups having significant relations based on Gene Set Enrichment Analysis, while detecting only 42 such communities in single layer network, has been a sign of progress
- Keywords:
- Data Analysis ; Biological Networks ; Community Detection ; Machine Learning ; Background Correction ; Chromatin Interaction Network ; Hi-C Technology
- محتواي کتاب
- view
- فهرست تصاویر
- فهرست جداول
- مقدمه
- پژوهشهای پیشین
- راهکار پیشنهادی
- نتایج ارزیابی
- جمعبندی و کارهای آتی
- مراجع