Transcript Slides - Department of Computer Science

CS 6243 Machine Learning
Advanced topic: pattern
recognition (DNA motif finding)
Final Project
• Draft description available on course website
• More details will be posted soon
• Group size 2 to 4 acceptable, with higher
expectation for larger teams
• Predict Protein-DNA binding
Biological background for TF-DNA binding
Genome is fixed – Cells are
dynamic
• A genome is static
– (almost) Every cell in our body has a copy of
the same genome
• A cell is dynamic
– Responds to internal/external conditions
– Most cells follow a cell cycle of division
– Cells differentiate during development
Gene regulation
• … is responsible for the dynamic cell
• Gene expression (production of protein) varies
according to:
–
–
–
–
–
Cell type
Cell cycle
External conditions
Location
Etc.
Where gene regulation takes place
• Opening of chromatin
• Transcription
• Translation
• Protein stability
• Protein modifications
Transcriptional Regulation of genes
Transcription Factor (TF)
(Protein)
RNA polymerase
(Protein)
DNA
Promoter
Gene
Transcriptional Regulation of genes
Transcription Factor (TF)
(Protein)
RNA polymerase
(Protein)
DNA
TF binding site, cis-regulatory element
Gene
Transcriptional Regulation of genes
Transcription Factor
(Protein)
RNA polymerase
DNA
TF binding site, cis-regulatory element
Gene
Transcriptional Regulation of genes
New protein
RNA
polymerase
Transcription Factor
DNA
TF binding site, cis-regulatory element
Gene
The Cell as a Regulatory Network
If C then D
gene D
A
B
C
Make D
If B then NOT D
If A and B then D
D
gene B
D
C
Make B
If D then B
Transcription Factors Binding to DNA
Transcriptional regulation:
• Transcription factors
bind to DNA
Binding recognizes
specific DNA
substrings:
• Regulatory motifs
Experimental methods
• DNase footprinting
– Tedious
– Time-consuming
• High-throughput techniques: ChIP-chip, ChIPseq
– Expensive
– Other limitations
Protein Binding Microarray
Computational methods for finding
cis-regulatory motifs
.
.
.
Given a collection of genes that are believed to be
regulated by the same/similar protein
– Co-expressed genes
– Evolutionarily conserved genes
Find the common TF-binding motif from promoters
Essentially a Multiple Local
Alignment
instance
.
.
.
• Find “best” multiple local alignment
• Multidimensional Dynamic Programming?
– Heuristics must be used
Characteristics of cis-Regulatory
Motifs
• Tiny (6-12bp)
• Intergenic regions are
very long
• Highly Variable
• ~Constant Size
– Because a constant-size
transcription factor binds
• Often repeated
• Often conserved
Motif representation
• Collection of exact words
– {ACGTTAC, ACGCTAC, AGGTGAC, …}
• Consensus sequence (with wild cards)
– {AcGTgTtAC}
– {ASGTKTKAC} S=C/G, K=G/T (IUPAC code)
• Position-specific weight matrices (PWM)
Position-Specific Weight Matrix
1
2
3
4
5
6
7
8
9
A
.97
.10
.02
.03
.10
.01
.05
.85
.03
C
.01
.40
.01
.04
.05
.01
.05
.05
.03
G
.01
.40
.95
.03
.40
.01
.3
.05
.03
T
.01
.10
.02
.90
.45
.97
.6
.05
.91
A
S
G
T
K
T
K
A
C
frequency
Sequence Logo
1
2
3
4
5
6
7
8
9
A
.97
.10
.02
.03
.10
.01
.05
.85
.03
C
.01
.40
.01
.04
.05
.01
.05
.05
.03
G
.01
.40
.95
.03
.40
.01
.3
.05
.03
T
.01
.10
.02
.90
.45
.97
.6
.05
.91
http://weblogo.berkeley.edu/
http://biodev.hgen.pitt.edu/cgi-bin/enologos/enologos.cgi
Sequence Logo
1
2
3
4
5
6
7
8
9
A
.97
.10
.02
.03
.10
.01
.05
.85
.03
C
.01
.40
.01
.04
.05
.01
.05
.05
.03
G
.01
.40
.95
.03
.40
.01
.3
.05
.03
T
.01
.10
.02
.90
.45
.97
.6
.05
.91
http://weblogo.berkeley.edu/
http://biodev.hgen.pitt.edu/cgi-bin/enologos/enologos.cgi
Entropy and information content
• Entropy: a measure of uncertainty
• The entropy of a random variable X that
can assume the n different values x1, x2, . .
. , xn with the respective probabilities p1,
p2, . . . , pn is defined as
Entropy and information content
• Example: A,C,G,T with equal probability
 H = 4 * (-0.25 log2 0.25) = log2 4 = 2 bits
 Need 2 bits to encode (e.g. 00 = A, 01 = C, 10 = G, 11 = T)
 Maximum uncertainty
• 50% A and 50% C:
 H = 2 * (-0. 5 log2 0.5) = log2 2 = 1 bit
• 100% A
 H = 1 * (-1 log2 1) = 0 bit
 Minimum uncertainty
• Information: the opposite of uncertainty
 I = maximum uncertainty – entropy
 The above examples provide 0, 1, and 2 bits of information,
respectively
Entropy and information content
A
C
G
T
1
2
3
4
5
6
7
8
9
.97
.01
.01
.01
.10
.40
.40
.10
.02
.01
.95
.02
.03
.04
.03
.90
.10
.05
.40
.45
.01
.01
.01
.97
.05
.05
.3
.6
.85
.05
.05
.05
.03
.03
.03
.91
H
I
.24 1.72 .36 .63 1.60 0.24 1.40 0.85 0.58
1.76 0.28 1.64 1.37 0.40 1.76 0.60 1.15 1.42
Mean
1.15
Total
10.4
Expected occurrence in random DNA: 1 / 210.4 = 1 / 1340
Expected occurrence of an exact 5-mer: 1 / 210 = 1 /
Sequence Logo
1
2
3
4
5
6
7
8
9
A
C
.97
.10
.02
.03
.10
.01
.05
.85
.03
.01
.40
.01
.04
.05
.01
.05
.05
.03
G
T
I
.01
.40
.95
.03
.40
.01
.3
.05
.03
.01
.10
.02
.90
.45
.97
.6
.05
.91
1.76 0.28 1.64 1.37 0.40 1.76 0.60 1.15 1.42
Real example
• E. coli. Promoter
• “TATA-Box” ~ 10bp upstream of transcription
start
• TACGAT
• TAAAAT
• TATACT
Consensus: TATAAT
• GATAAT
• TATGAT
Note: none of the instances
• TATGTT
matches the consensus perfectly
Finding Motifs
Classification of approaches
• Combinatorial algorithms
– Based on enumeration of words and
computing word similarities
• Probabilistic algorithms
– Construct probabilistic models to distinguish
motifs vs non-motifs
Combinatorial motif finding
• Idea 1: find all k-mers that appeared at least m times
– m may be chosen such that # occurrence is statistically
significant
– Problem: most motifs have divergence. Each variation may only
appear once.
• Idea 2: find all k-mers, considering IUPAC nucleic acid
codes
– e.g. ASGTKTKAC, S = C/G, K = G/T
– Still inflexible
• Idea 3: find k-mers that approximately appeared at least
m times
– i.e. allow some mismatches
Combinatorial motif finding
Given a set of sequences S = {x1, …, xn}
• A motif W is a consensus string w1…wK
• Find motif W* with “best” match to x1, …, xn
Definition of “best”:
d(W, xi) = min hamming dist. between W and a word in xi
d(W, S) = i d(W, xi)
W* = argmin( d(W, S) )
Exhaustive searches
1. Pattern-driven algorithm:
For W = AA…A to TT…T
(4K possibilities)
Find d( W, S )
Report W* = argmin( d(W, S) )
Running time: O( K N 4K )
(where N = i |xi|)
Guaranteed to find the optimal solution.
Exhaustive searches
2. Sample-driven algorithm:
For W = a K-char word in some xi
Find d( W, S )
Report W* = argmin( d( W, S ) )
OR Report a local improvement of W*
Running time: O( K N2 )
Exhaustive searches
• Problem with sample-driven approach:
• If:
– True motif does not occur in data, and
– True motif is “weak”
• Then,
– random strings may score better than any
instance of true motif
Example
• E. coli. Promoter
• “TATA-Box” ~ 10bp upstream of transcription
start
• TACGAT
• TAAAAT
• TATACT
Consensus: TATAAT
• GATAAT
Each instance differs at most 2
• TATGAT
bases from the consensus
• TATGTT
None of the instances matches the
consensus perfectly
Heuristic methods
• Cannot afford exhaustive search on all
patterns
• Sample-driven approaches may miss real
patterns
• However, a real pattern should not differ
too much from its instances in S
• Start from the space of all words in S,
extend to the space with real patterns
Some of the popular tools
• Consensus (Hertz & Stormo, 1999)
• WINNOWER (Pevzner & Sze, 2000)
• MULTIPROFILER (Keich & Pevzner,
2002)
• PROJECTION (Buhler & Tompa, 2001)
• WEEDER (Pavesi et. al. 2001)
• And dozens of others
Probabilistic modeling approaches
for motif finding
Probabilistic modeling approaches
• A motif model
– Usually a PWM
– M = (Pij), i = 1..4, j = 1..k, k: motif length
• A background model
– Usually the distribution of base frequencies in
the genome (or other selected subsets of
sequences)
– B = (bi), i = 1..4
• A word can be generated by M or B
Expectation-Maximization
• For any word W,
 P(W | M) = PW[1] 1 PW[2] 2…PW[K] K
 P(W | B) = bW[1] bW[2] …bW[K]
• Let  = P(M), i.e., the probability for any word to
be generated by M.
• Then P(B) = 1 - 
• Can compute the posterior probability P(M|W)
and P(B|W)
 P(M|W) ~ P(W|M) * 
 P(B|W) ~ P(W|B) * (1-)
Expectation-Maximization
Initialize:
Randomly assign each word to M or B
• Let Zxy = 1 if position y in sequence x is a motif, and 0
otherwise
• Estimate parameters M, , B
Iterate until converge:
• E-step: Zxy = P(M | X[y..y+k-1]) for all x and y
• M-step: re-estimate M,  given Z (B usually fixed)
Expectation-Maximization
position
5
1
Initialize
E-step
probability
1
5
9
9
M-step
• E-step: Zxy = P(M | X[y..y+k-1]) for all x and y
• M-step: re-estimate M,  given Z
MEME
•
•
•
•
•
Multiple EM for Motif Elicitation
Bailey and Elkan, UCSD
http://meme.sdsc.edu/
Multiple starting points
Multiple modes: ZOOPS, OOPS, TCM
Gibbs Sampling
• Another very useful technique for
estimating missing parameters
• EM is deterministic
– Often trapped by local optima
• Gibbs sampling: stochastic behavior to
avoid local optima
Gibbs Sampling
Initialize:
Randomly assign each word to M or B
• Let Zxy = 1 if position y in sequence x is a motif, and 0
otherwise
• Estimate parameters M, B, 
Iterate:
•
•
•
•
•
Randomly remove a sequence X* from S
Recalculate model parameters using S \ X*
Compute Zx*y for X*
Sample a y* from Zx*y.
Let Zx*y = 1 for y = y* and 0 otherwise
Gibbs Sampling
probability
position
0.2
probability
0.15
0.1
0.05
0
0
2
4
6
8
10
position
12
14
16
18
Sampling
• Gibbs sampling: sample one position according to probability
•
•
– Update prediction of one training sequence at a time
Viterbi: always take the highest
Simultaneously update
EM: take weighted average
predictions of all sequences
20
Better background model
• Repeat DNA can be confused as motif
– Especially low-complexity CACACA… AAAAA, etc.
• Solution: more elaborate background model
– Higher-order Markov model
0th order: B = { pA, pC, pG, pT }
1st order: B = { P(A|A), P(A|C), …, P(T|T) }
…
Kth order: B = { P(X | b1…bK); X, bi{A,C,G,T} }
Has been applied to EM and Gibbs (up to 3rd order)
Gibbs sampling motif finders
• Gibbs Sampler
– First appeared as: Larence et.al. Science 262(5131):208-214.
– Continually developed and updated. webpage
– The newest version: Thompson et. al. Nucleic Acids Res. 35 (s2):W232W237
• AlignACE
– Hughes et al., J. of Mol Bio, 2000 10;296(5):1205-14.
– Allow don’t care positions
– Additional tools to scan motifs on new seqs, and to compare and group
motifs
• BioProspector, X. Liu et. al. PSB 2001 , an improvement of
AlignACE
– Liu, Brutlag and Liu. Pac Symp Biocomput. 2001;:127-38.
– Allow two-block motifs
– Consider higher-order markov models