Comparative sequence analysis for ENCODE functional elements

1
Comparative sequence analysis for ENCODE
functional elements

• Shamil Sunyaev
• Genetics Division, Brigham Womens Hospital,
• Harvard Medical School
• Harvard-M.I.T. Division of HST

2
Features of a good score

• Reflects phylogeny
• Based on a good model of evolution/mutation
• Robust to missing information

3
General strategy

• Reject the null model of neutrality
• Develop a score that partitions columns based on
  deviation from neutrality
• Compute a p-value for neutrality of site based
  on heuristic score

4
Probabilistic conservation score

• The score is position specific
• The score is based on a complex mutation model
• The score is expressed as a p-value that a given
  nucleotide position evolves neutrally

5
Determining mutation rates through phylogenetic
inference
human
GATCTATGGTGTGGGTTTGCATAACAGGAGAGGA

chimp
GATCCATGGTGTGGGTTTGCATAATAGGAGAGGA
baboon
GTTCCATGGTGTGGGTTTGCATAATAGGGGAGGA
Probability of T?C1/10 Probability of C?T1/3
6
Human Chimp Baboon
AGC
ACC
AGC
Mutation rates are modeled as asymmetric and
context specific. The model incorporates
insertions and deletions
AGC
G?C

• Scores
• Empirical Score ln(lcons/lmut)
• ML estimate of fixation probability
• Log-likelihood ratio Score ln(lML/lNeutral)

P-value
7
Extending substitution rates to other species
Instantaneous rate matrix of transitions Q
P(t) eQt

• Ignores mutation rate heterogeneity
• Assumes uniformity between species

8
Computing a heuristic score
G
G
T
G
T
9
Computing a p-value
G
C
T
C
T
P(X?S) n(xltS)/N
10
Identifying blocks of conserved positions
We can transform a site-specific score into a
method for identifying conserved regions
Si ?j0..i -ln(pj) ln(t)
11
(No Transcript)
12
Analysis of ENCODE tracks

• Distributions of the score for individual
  positions
• Element- and Region-specific p-values
• Overlaps of functional regions with highly
  conserved regions (also for PhastCons, BinCons
  and GERP)
• Length/conservation distribution
• Conservation only in a subset of phylogeny
• Alignability of regions

13
Analysis of ENCODE tracks

• Distributions of the score for individual
  positions
• Element- and Region-specific p-values
• Overlaps of functional regions with highly
  conserved regions (also for PhastCons, BinCons
  and GERP)
• Length/conservation distribution
• Conservation only in a subset of phylogeny
• Alignability of regions

14
(No Transcript)
15
Analysis of ENCODE tracks

• Distributions of the score for individual
  positions
• Element- and Region-specific p-values
• Overlaps of functional regions with highly
  conserved regions (also for PhastCons, BinCons
  and GERP)
• Length/conservation distribution
• Conservation only in a subset of phylogeny
• Alignability of regions

16
Additive score for the region
Expect normal distribution
P-value
17
Analysis of ENCODE tracks

• Distributions of the score for individual
  positions
• Element- and Region-specific p-values
• Overlaps of functional regions with highly
  conserved regions (also for PhastCons, BinCons
  and GERP)
• Length/conservation distribution
• Conservation only in a subset of phylogeny
• Alignability of regions

18
Overlaps
19
Analysis of ENCODE tracks

• Distributions of the score for individual
  positions
• Element- and Region-specific p-values
• Overlaps of functional regions with highly
  conserved regions (also for PhastCons, BinCons
  and GERP)
• Length/conservation distribution
• Conservation only in a subset of phylogeny
• Alignability of regions

20
Score
Length of region
21
Analysis of ENCODE tracks

• Distributions of the score for individual
  positions
• Element- and Region-specific p-values
• Overlaps of functional regions with highly
  conserved regions (also for PhastCons, BinCons
  and GERP)
• Length/conservation distribution
• Conservation only in a subset of phylogeny
• Alignability of regions

22
Analysis of ENCODE tracks

• Distributions of the score for individual
  positions
• Region-specific p-values
• Overlaps of functional regions with conserved
  regions (also for PhasCons, BinCons and GERP
• Length/conservation distribution
• Conservation only in a subset of phylogeny
• Alignability of regions