Computational Genomics and Proteomics

1
Computational Genomics and Proteomics
Lecture 8 Motif Discovery
2
Outline Gene Regulation DNA Transcription
factors Motifs What are they? Binding
Sites Combinatoric Approaches Exhaustive
searches Consensus Comparative
Genomics Example Probabilistic
Approaches Statistics EM algorithm Gibbs
Sampling
3
www.accessexcellence.org
4
www.accessexcellence.org
5
www.accessexcellence.org
6
Four DNA nucleotide building blocks
G-C is more strongly hydrogen-bonded than A-T
7
Degenerate code
Four bases A, C, G, T Two-fold degenerate IUB
codes RAG -- Purines YCT --
Pyrimidines KGT MAC SGC WAT Four-fold
degenerate NAGCT
8
Transcription Factors

• Required but not a part of the RNA polymerase
  complex
• Many different roles in gene regulation
• Binding
• Interaction
• Initiation
• Enhancing
• Repressing
• Various structural classes (eg. zinc finger
  domains)
• Consist of both a DNA-binding domain and an
  interactive domain

9
Motifs

• Short sequences of DNA or RNA (or amino acids)
• Often consist of 5- 16 nucleotides
• May contain gaps
• Examples include
• Splice sites
• Start/stop codons
• Transmembrane domains
• Centromeres
• Phosphorylation sites
• Coiled-coil domains
• Transcription factor binding sites (TFBS
  regulatory motifs)

10
TFBSs

• Difficult to identify
• Each transcription factor may have more than one
  binding site
• Degenerate
• Most occur upstream of translation start site
  (TSS) but are known to also occur in
• introns
• exons
• 3 UTRs
• Usually occur in clusters, i.e. collections of
  sites within a region (modules)
• Often repeated
• Sites can be experimentally verified

11
Why are TFBSs important?

• Aid in identification of gene networks/pathways
• Determine correct network structure
• Drug discovery
• Switch production of gene product on/off

Gene A Gene B
12
Consensus sequences

• Matches all of the example sequences closely but
  not exactly
• A single site
• TACGAT
• A set of sites
• TACGAT
• TATAAT
• TATAAT
• GATACT
• TATGAT
• TATGTT
• Consensus sequence
• TATAAT or
• TATRNT
• Trade-off number of mismatches allowed,
  ambiguity in consensus sequence and the
  sensitivity and precision of the representation.

13
Information Content and Entropy
14
Sequence Logos
15
Frequency Matrices

• Given a collection of motifs,
• TACGAT
• TATAAT
• TATAAT
• GATACT
• TATGAT
• TATGTT
• Create the matrix

T A C G
16
Position weight matrices
17
Finding Motifs

• Two problems
• Given a collection of known motifs, develop a
  representation of the motifs such that additional
  occurrences can reliably be identified in new
  promoter regions
• Given a collection of genes, thought to be
  related somehow, find the location of the motif
  common to all and a representation for it.
• Two approaches
• Combinatorial
• Probabilistic

18
Combinatorial Approach
19
Exhaustive Search
20
Exhaustive Search
Sample-driven here refers to trying all the words
as they occur in the sequences, instead of trying
all possible (4W) words exhaustively
21
Greedy Motif Clustering
22
Greedy Motif Clustering
23
Greedy Motif Clustering
24
Comparative Genomics

• Main Idea Conserved non coding regions are
  important
• Align the promoters of orthologous co-expressed
  genes from two (or more) species e.g. human and
  mouse
• Search for TFBS only in conserved regions
• Problems
• Not all regulatory regions are conserved
• Which genomes to use?

25
Phylogenetic Footprinting
Phylogenetic Footprinting refers to the task of
finding conserved motifs across different
species. Common ancestry and selection on these
motifs has resulted in these footprints.
26
Phylogenetic Footprinting An Example

• Xie et al. 2005
• Genome-wide alignments for four species (human,
  mouse, rat, dog)
• Promoter regions and 3UTRs then extracted for
  17,700 well-annotated genes
• Promoter region taken to be (-2000, 2000)
• This set of sequences then searched exhaustively
  for motifs

Nature 434, 338-345, 2005
27
The Search
Xie et al. 2005
28
Expected Rate
29
Probabilistic Approach
30
Gibbs Sampling (applied to Motif Finding)
31
Gibbs Sampling Algorithm
32
Gibbs Sampling Motif Positions
33
AlignACE - Gibbs Sampling
34
Remainder of the lecture Maximum likelihood and
the EM algorithm
The remaining slides are for your information
only and will not be part of the exam
35
Basic Statistics
36
Maximum Likelihood Estimates
37
EM Algorithm
38
Basic idea (MEME)
http//meme.nbcr.net/meme/meme-intro.html
39
Basic idea (MEME)
MEME is a tool for discovering motifs in a group
of related DNA or protein sequences. A motif is a
sequence pattern that occurs repeatedly in a
group of related protein or DNA sequences. MEME
represents motifs as position-dependent
letter-probability matrices which describe the
probability of each possible letter at each
position in the pattern. Individual MEME motifs
do not contain gaps. Patterns with
variable-length gaps are split by MEME into two
or more separate motifs. MEME takes as input a
group of DNA or protein sequences (the training
set) and outputs as many motifs as requested.
MEME uses statistical modeling techniques to
automatically choose the best width, number of
occurrences, and description for each motif.
http//meme.nbcr.net/meme/meme-intro.html
40
Basic MEME Model
41
MEME Background frequencies
42
MEME Hidden Variable
43
MEME Conditional Likelihood
44
EM algorithm
45
Example
46
E-step of EM algorithm
47
Example
48
M-step of EM Algorithm
49
Example
50
Characteristics of EM
51
Gibbs Sampling (versus EM)