Sequence Evolution Modeling

1
Sequence Evolution Modeling

• Programming Project
• SoCalBSI 2006
• Sarah Aerni

2
Evolution Theory

• Organisms share common ancestor
• At speciation, both genomes are identical
• Mutations accumulate over multiple generations
  causing organismal genomes to look dissimilar
• Some regions are conserved across organisms
• Genes
• Functional regulatory elements
• Regions of unknown function
• Distance between organisms can be measured by
  edit distance between genomes

3
Evolutionary Tree Reconstruction - History

• Emile Zuckerkandl and Linus Pauling pioneer
  reconstruction using DNA analysis
• Use parsimony to determine the shortest distance
  between any two genomes
• Used to study human populations
• Out of Africa Theory
• Modern humans did not evolve from Neanderthals
• Based on mitochondrial data
• Divergence between Neanderthals and humans is
  greater than the number of generations based on
  carbon dating
• Can be important in studying disease

4
Evolution of a Genome

• At speciation both sequences are identical
• Mutations occur in genomes separately
• Natural selection determines which changes will
  be carried on to future generations
• Gene duplication
• Positive selection increased fitness as a result
  of increased gene product (globin)
• Negative selection over-expression of genes
  cause defects (Downs Syndrome)
• Regulatory mutations
• Changes in transcription factor binding sites
  affects expression of genes

5
Simulating Sequence Evolution

• Build a model (markov 6) of the human genome
• Create a dictionary of all 7mers and record their
  corresponding frequencies in the genome
• Use model to create an initial sequence
• Assume all bases are represented equally (25
  chance of A,C,G or T) and randomly create a 6mer
• Elongate the sequence using probabilities given
  from the model built from real sequence data
• Assign regions of sequence to be of certain types
• Assign conserved (pick location, extend
  150-500bp)
• Determine where motifs will be located (regions
  will be considered conserved as long as motif is
  not destroyed)

6
Simulating Sequence Evolution

• Permit rounds of mutations to occur
• Mutations rate will based on the profile which is
  input by user
• Multiple motifs are possible
• Each location is required to have a nonzero
  probability of being seen and must meet input
  threshold

7
Simulating Sequence Evolution

• Mutation types are variable
• Point mutations and indels are represented
• Insertions will be repeats of the 1-3 base
  preceding or following the insertion location

8
Simulating Sequence Evolution

• Motif changes
• Motif will be chosen based on a profile and
  locations will be identified
• Mutation rates will be adjusted accordingly
• Rates will vary by location in the motif
• Loss and appearance of motifs will be adjusted
  dynamically (after each generation)
• If all motifs in a sequence are destroyed, the
  organism is reset to the previous generation
• Possible directions
• Motif finders can be used to re-identify the
  motifs
• Allow multiple motifs to be inserted
• Spontaneously pick up any motifs that match a
  profile
• Incorporate distance to start site?

9
Human-Rodent precursor model

• Peroxisome-proliferator-activated receptor gamma
• Bound by heterodimer
• Profile at right used for mutations in motifs

10
Human-Rodent precursor model

• 10 million generations
• 60 kb regions, 5 conservation initially
• Requires minimum 1 motif per sequence
• Threshold for motif is 80 identity
• Mutation rates
• Conserved 1 mutation/100mb/generation
• Non-conserved 1 mutation/60mb/generation
• 50 simulations to collect results

11
Human-Rodent results

• Distribution of mutations

• Average sequence size is 60041 bp, 5.02
  conserved after simulation
• pipmatcher program was able to pick up
  conservation regions
• 46 sequences had only one motif remaining, 3
  sequences with two, 1 sequence with three
• Would incorporation of selective advantage for
  multiple TFBS may change results?

12
Conservation results

• Used pipmaker
• Focus on exons still picks up conservation
• Strong argument that conserved regions contain
  elements of interest

13
Conservation results
14
Motif Results

• Most sequences experience at least one motif loss
• They occur indiscriminatelythroughout
  simulation

15
Motif Results

• Most sequences do not evolve multiple binding
  sites
• Death of a motif is more common
• Adjust model for improved fitness?

16
Motif results

• Identified motif locations were used to create a
  new logo
• Logo is very similar to actual logo
• If threshold is increased, profile would look
  more strict
• Could this technique be used to determine motif
  finding thresholds?

Logo created from motifs in evolved sequences
Original logo
17
Further development?

• Try to answer more complex evolutionary events
• Gene duplication
• Incorporate positive and negative selection for
  binding sites
• Incorporate multiple motifs
• Try to rediscover sites with motif finders
• Model could easily be used as a source of
  synthetic data for training motif finders
• Determine appropriate thresholds

18
Acknowledgments and References

• Ali Mortazavi
• SoCalBSI
• Wold Lab
• http//weblogo.berkeley.edu
• http//pipmaker.bx.psu.edu
• JASPAR database
• Cistematic