1. Gene regulatory network
Jin Chen
CSE
2012 Fall

2. Outline Transcriptional regulation Co-expression & co-regulation
Bio-techniques
ChIP-seq
Bacterial one-hybrid system
Computational models for gene regulation network construction
Binding + expressions with TF knock out
Binding + time serial gene expressions

3. David J.C. MacKay, Information Theory, Inference & Learning Algorithms, 2003

4. Transcriptional regulation
Regulation of transcription controls when transcription occurs and how much RNA is created
Transcription factors are often needed to be bound to a regulatory binding site to switch a gene on (activator) or to shut off a gene (repressor)
Generally, as the organism grows more sophisticated, their cellular protein regulation becomes more complicated
Example: when a cell contains a surplus amount of the amino acid tryptophan, the acid binds to a specialized repressor protein (tryptophan repressor). The binding changes the structural conformity of the repressor such that it binds to the operator region for the operon that synthesizes tryptophan, preventing their expression and thus suspending production. This is a form of negative feedback.
and indeed some human genes can be controlled by many activators and repressors working together

5. Transcription control
Transcription control is directed primarily by two elements
Transcription factors (TF)
DNA sequences that facilitate the binding of these TFs (cis-regulatory elements)

6. Transcription control
First, there needs to be an initiating signal
This signal gives rise to the activation of a TF, and recruits other members of the "transcription machine." TFs generally simultaneously bind DNA. TFs and their cofactors, can be regulated through reversible structural alterations
Transcription is initiated at the promoter site, as an increase in the amount of an active TF binds a target DNA sequence. Other proteins, known as "scaffolding proteins" bind other cofactors and hold them in place
Frequently, extracellular signals induce the expression of immediate early genes. These are in and of themselves TFs or components thereof, and can further influence gene expression
In order for a gene to be expressed, several things must happen.

7. Gene regulation network
Node
TF or target gene
Edge
Regulation relation
Directed
Activation “--->”
Inhibition “---|”
Yeast
Mata et al. Genome Biol. 2007;8(10):R217

8. Co-expression & co-regulation
Genes belonging to the same cluster are often called co-expressed
Genes with similar expression patterns might share transcription factors and functional regulatory binding sites

9. Topic 1. GRN Reconstruction
Background
Topic 1. GRN Reconstruction
Algorithms for GRN reconstruction based on gene expression & motif data ( )
Algorithms focusing on integrating binding data with existing data (2007-now)
Time
2002
2008
2007
2010
2 input data
Time serial microarray
TF binding motifs
3 models
Time shift matching
Mutual information
Granger causality & DBN
2 limitations
High false positive rate
Small scale (# genes~100)
3 input data
Time serial microarray
TF binding motifs
Binding data
2 models
Time shift matching and binding
Protein expression approximation from binding data
2 limitations
Lack of binding data at systems level
Combinatorial TF studies
To see the interaction among regulatory genes, people utilize large-scale perturbation assays in which expression of each gene was interrupted.
A putative GRN built from time serial gene expression data or binding data may have a high proportion of false positives. Due to the algorithm scalability and over-fitting problems of existing learning systems, the direct purification of a large GRN is very difficult.

10. ChIP-seq
ChIP-seq is the sequencing of the genomic DNA fragments that co-precipitate with a DNA-binding protein that is under study
DNA-binding proteins most frequently investigated in this way are transcription factors, etc
ChIP-seq can identify all DNA segments in the genome physically associated with a specific DNA-binding protein
It does not rely on prior knowledge of precise DNA binding sites
precise mapping of DNA-binding proteins/complexes
transcription factors, chromatin-modifying enzymes, modified histones interacting with genomic DNA, and components of the basal transcriptional machinery
Liu et al, BMC Biology 2010, 8:56

11. Flow scheme of the central steps in the ChIP-seq procedure
Liu et al. BMC Biology :56  

12. Example Algorithmic analysis for mapping and peak-calling
Helicos DNA Sample Preparation Methodology for ChIP-Seq
A ChIP DNA sample from a stem cell population and the corresponding input DNA sample were both processed without amplification
A ChIP DNA sample from a stem cell population and the corresponding input DNA sample were both processed without amplification

13. Three key steps in ChIP-seq
Antibody selection
Sequencing
Algorithmic analysis for mapping and peak-calling

14. Binding network
First action of a TF is to find and to bind DNA segments and ChIP-seq allows the binding sites of TFs to be identified across entire genomes
Protein-DNA binding network
Direct downstream targets of any transcription factor can be determined
DNA sequence motif that is recognized by the binding protein can be computed
Precise regulatory sites in the genome for any transcription factor can be identified
Clustering of transcription-regulatory proteins at specific DNA sites can be assessed

15. Example
ChIP-seq profiling of 13 TFs in embryonic stem (ES) cell development revealed the organization of regulatory elements. This provided insights in the integration of TF-mediated signaling pathways in ES cell differentiation
Chen et al Cell 2008 , 133:

16. What ChIP-seq cannot do
Many observed binding events are neutral and do not regulate transcription
Regulatory binding events often occur at enhancers that are not proximal to the target gene that they control
 The task of identifying transcriptional targets requires the integration of ChIP-seq with evidence from expression data to help associate binding events with target gene regulation
Honkela et al. PNAS 2010 vol. 107 no. 17 pp 7793–7798

17. Bacterial one-hybrid system
wikipedia

18. Gene expression data TF knock-out gene expression
TF over-expression gene expression
Differential expression of genes between wild type and mutant/over-expression is indicative of a potential regulatory interaction, e.g. Yeast GRN
Reimand et al, Nucleic Acids Research, 2010, Vol. 38, No. 14 pp 4768–4777

19. Comprehensive analysis of TF knockout expression data in Yeast
269 TF knockout microarrays, covering almost all yeast TFs
DNA–protein interactions derived from ChIP-chip experiments
Predicted TF binding sites with position weight matrices
Hu et al. Nat. Genet., 39, 683–687, 2007; Harbison et al. Nature, 431, 99–

20. Comprehensive analysis of TF knockout expression data in Yeast
Checked the expression levels of the TFs
Intuitively one expects the TF under consideration to have lower expression in the mutant strain compared with the wild type strain
confirms this for 155 TFs
78 TFs display a negative fold change at statistically non-significant levels
36 TFs are lethal
Among these regulators are several that affect many genes, emphasising that even small adjustments to TF expression levels can have a dramatic effect on target genes.
Reimand et al, Nucleic Acids Research, 2010, Vol. 38, No. 14 pp 4768–4777

21. Reimand et al, Nucleic Acids Research, 2010, Vol. 38, No
Reimand et al, Nucleic Acids Research, 2010, Vol. 38, No. 14 pp 4768–4777

22. Comprehensive analysis of TF knockout expression data in Yeast
Examine functional annotations of differentially expressed genes
As most TFs are considered to regulate distinct cellular processes, their target genes should be associated with a coherent set of molecular and biological functions
Used g:Profiler to identify GO, KEGG and Reactome pathway annotations
Among these regulators are several that affect many genes, emphasising that even small adjustments to TF expression levels can have a dramatic effect on target genes.
Across all TF knockouts, this analysis has a higher score than the original analysis
Reimand et al, Nucleic Acids Research, 2010, Vol. 38, No. 14 pp 4768–4777

23. Among these regulators are several that affect many genes, emphasising that even small adjustments to TF expression levels can have a dramatic effect on target genes.
Reimand et al, Nucleic Acids Research, 2010, Vol. 38, No. 14 pp 4768–4777

24. Comprehensive analysis of TF knockout expression data in Yeast
Overlap between TF-binding and TF knockout data
Collect binding sites for 142 TFs, comprising 5,188 ChIP-chip interactions and 17,091 motif predictions
Calculate the intersection between the list of differentially expressed genes from the TF knockout and targets identified by ChIP-chip or binding-site predictions
2,230 regulation relations
Among these regulators are several that affect many genes, emphasising that even small adjustments to TF expression levels can have a dramatic effect on target genes.
Information from large-scale ChIP-chip and motif-finding studies were considered
Reimand et al, Nucleic Acids Research, 2010, Vol. 38, No. 14 pp 4768–4777

25. Comprehensive analysis of TF knockout expression data in Yeast
Include protein-protein interaction information as an additional perspective to the assessment of GRN
TFs that function together may show significant overlap in their target genes
Out of 115 pairs of physically interacting TFs in the dataset, 92 display such an overlap
TFs tend to regulate genes that interact with each other
Out of 110,487 differentially expressed genes, there are 3,846 pair-wise interactions between co-regulated genes, covering 2,262 genes in total
Most TFs target at least one pair of interacting genes
Among these regulators are several that affect many genes, emphasising that even small adjustments to TF expression levels can have a dramatic effect on target genes.
Reimand et al, Nucleic Acids Research, 2010, Vol. 38, No. 14 pp 4768–4777

26. Temporal gene expression data
A problem with the above approach is that the creation of mutant strains is challenging or impossible for many TFs of interest
Even when available, mutants may provide very limited information because of redundancy or due to the confounding of signal from indirect regulatory feedback
Temporal dynamics: use time serial wild-type gene expression. e.g. Drosophila GRN
Honkela et al. PNAS 2010 vol. 107 no. 17 pp 7793–7798

27. Other models for gene regulation network construction
Expression based study
Dynamic Bayesian Network
Granger causality
TF binding motif based study
Weader
AlignAce

28. Gene regulation network analysis
Transcriptional regulation is mediated by the combinatorial interplay between cis-regulatory DNA elements and trans-acting transcription factors, and is perhaps the most important mechanism for controlling gene expression
A transcriptional regulatory network that integrates such information can lead to a systems-level understanding of regulatory mechanisms
Kim et al. Wiley Interdisciplinary Reviews: Systems Biology and Medicine. Vol. 3, Iss 1, pp 21–

29. Discovery of motifs and regulatory modules
Binding motif prediction only based on the knowledge afforded by PWMs often suffers from high false-positive rates
To reduce the false-positive rate, a number of biological insights have been used
evolutionary pressure placed on these important cis-regulatory elements
co-expression with genes that have well-documented functions and expression patterns
clustering of cis-regulatory features into cis-regulatory modules

30. Discovery of motifs and regulatory modules
Collect the binding sequences for known TFs and to identify potential binding sites in unannotated genome sequence
position frequency matrix
cluster of individual TF binding sites
regulatory relations among 4 genes
As the majority of transcription factors are sequence-specific, there have been extensive efforts
Modeling of individual transcription factor binding sites and cis-regulatory modules
Kim et al. Wiley Interdisciplinary Reviews: Systems Biology and Medicine. Vol. 3, Iss 1, pp 21–

31. Co-regulation
A density-based subspace clustering algorithms for coherent clustering of gene expression data
The model allows
Expression profiles of genes in a cluster to follow any shifting-and-scaling patterns in subspace
Expression value changes across any two conditions of the cluster to be significant
Experimental results show that the algorithm is able to detect a significant amount of high biological significant clusters missed by previous models