ITEM METADATA RECORD
Title: IDENTIFICATION OF CIS-REGULATORY MODULES AND NON-CODING VARIATION USING MACHINE LEARNING METHODS
Other Titles: IDENTIFICATIE VAN CIS-REGULATORISCHE MODULES EN NIET-CODERENDE VARIATIE DOOR MIDDEL VAN MACHINE LEARNING METHODES
Authors: Svetlichnyy, Dmitry; R0363204
Issue Date: 24-Oct-2016
Table of Contents: Table of Contents
CHAPTER 1: INTRODUCTION 1
1. Transcriptional regulation 1
1.1 Classes of cis-regulatory modules 2
1.1.1 CRM architecture 3
1.2 Chromatin signatures of CRMs 4
1.3 Motif and evolutionary constraint in noncoding regions 5
1.4 Detecting regulatory regions using experimental methods 7
1.4.1 Genome-wide identification of TF binding with ChIP and DamID 7
1.4.2 Identification of enhancers using open chromatin profiling. 7
1.4.3 Functional validation of enhancers 8
1.4.3.1 Massively parallel reporter assay 8
1.4.2.2 STARR-seq 9
1.4.2.3 Assays using genomic integration 9
2. Computational identification of regulatory elements in the genome 9
2.1 Motif-based approaches 9
2.2 Comparative genomics approaches to identify functional binding sites 10
2.3 CRM detection using motif clustering 11
2.4 Machine learning approaches to find CRMs 12
2.4.1 Unsupervised learning methods 12
2.4.1.1 Hidden Markov Models 12
2.4.2 Supervised methods 13
2.4.2.1 Evaluation of model performance 13
2.4.2.2 Regularized linear models 14
2.4.2.3 SVM for CRM prediction 14
2.4.2.4 Ensemble of decision trees 15
2.4.2.4.1 Algorithms to train a decision tree classifier 16
2.4.2.4.2 Parameters of the Random Forest 17
2.4.2.5 Feature selection methods 17
2.4.2.5.1 Filter methods 18
2.4.2.5.2 Wrapper methods 18
2.4.2.5.3 Embedded methods 19
2.4.2.6 Deep learning methods 19
2.4.2.6.1 Convolutional Neural Networks 19
2.4.2.6.2 Overfitting in the CNN 21
2.4.2.6.3 CNNs for computational identification of CRMs 21
3. Transcriptional regulation and cancer 22
3.1 Role of TP53 in cancer 23
3.2 Role of non-coding mutations in cancer 23
CHAPTER II: Objectives 27
CHAPTER III: Results 29
PAPER 1: Identification of High-Impact cis-Regulatory Mutations Using Transcription Factor Specific Random Forest Models 31
PAPER 2: Multiplex enhancer-reporter assays uncover unsophisticated TP53 enhancer logic 83
CHAPTER IV: DISCUSSION 141
5.1 Computational models to identify TF-specific enhancers 141
5.2 Prediction of high-impact cis-regulatory mutations with enhancer models 142
5.3 Deciphering p53 enhancer logic using high-throughput enhancer reporter assays coupled with machine learning 143
5.4 General conclusion 146
5.5 Future perspectives 147
BIBLIOGRAPHY 153
Publication status: published
KU Leuven publication type: TH
Appears in Collections:Department of Human Genetics - miscellaneous
Laboratory of Computational Biology

Files in This Item:
File Status SizeFormat
Svetlichnyy_PhD_Thesis.pdf Published 29649KbAdobe PDFView/Open Request a copy

These files are only available to some KU Leuven Association staff members

 




All items in Lirias are protected by copyright, with all rights reserved.