Phylogenetic footprinting

1
Phylogenetic footprinting
Topics in Computational Biology Ilkka
Vaahtoranta 4.3.2004
2
Phylogenetic footprinting

• Introduction
• Methods used
• Substring Parsimony Problem (Torsten)
• Results

3
Introduction
4
The Problem

• Major challenge of current genomics is to
  understand how gene expression is regulated.
• An important step towards this understanding is
  the capability to identify regulatory elements.

5
In a Nutshell

• Phylogenetic footprinting is a method for the
  discovery of regulatory elements in a set of
  orthologous regulatory regions from multiple
  species. It does so by identifying the best
  conserved motifs in those orthologous regions.

• Idea of phylogenetic footprinting was first
  invented as early as 1988 (Tagle, Koop,
  Goodman...)
• It was at that time little ahead of its time
• Only few sequences from related species were
  available

6
Orthologous vs. Analogous

• Orthologous sequences have the same function in
  different species and are related
• Analogous sequences have same kind of function
  but are not related
• Phylogenetic footprinting uses othologous
  sequences

7
Regulatory Elements
RE's
Exon
Intron
5
Promoter sequence
Gene
Promoter

• Lies usually before the actual gene
• Rarely after the gene
• Aproximately 600-1000 bp long sequence
• Holds regulatory elements

Regulatory elements

• Relatively short sequences, from 5 to 25 bp long
• May hold gaps
• Appear in othervice non-functional sequence

8
Multiple genes in single species

• Single species
• Related genes
• This technique is used to find common regulatory
  factors
• Only in given organism
• REs of single gene are not found
• This is not phylogenetic footprinting

9
Multiple species with orthologous regulatory
regions
What do we need to identify regulatory elements?

• Set of orthologous non-functinal DNA from species
  that are related
• For an example one might use the non-coding
  sequence of insulin in ten different vertebrates
• If well conserved, possible RE
• This is phylogenetic footprinting

10
Why examine non-functional sequences?

• Functional sequences evolve slower rate than
  non-functional sequences cause of the selective
  pressure
• A transition in a functional sequence (gene) may
  change the whole function of coded protein
• A transition in a non-functional sequence (RE)
  may only change expression freqvency of a gene

11
Phylogenetic footprinting exploits the mutation
rate difference of functional and non-functional
sequences
12
Methods used
13
Global Multiple Alignment

• CLUSTLAW, GMA tool
• Global Multiple alignment drawbacks
• It is a np hard problem.
• If optimal MA could identify all REs, we could
  not compute it.
• Because REs are quite short (10 in 1000
  nucleotides), noise of diverged non-functional
  sequences will overcome the short conserved
  signal.

14
Classical motif finding

• Standard motif finding (MEME, AlignAce,
  ANN-Spec....)
• Segment based motif finding (DIAGLIN...)
• Outperform global multipple alignment

All have important shortcoming

• Do not take account phylogenetic relationships
• Closely related sequences have too high weight

15
Substring Parsimony Problem, a motif finding
algorithm

• Formalization of the PF idea
• Also NP-hard problem but easy to tune up to
  eliminate exponential behavior
• Substring parsimony searches for best alignments
  in given sequence set.
• Difference between substringparsimony and
  multiple alignment lies in given phylogenetic
  tree.
• Multiple alignment does not care about
  relationships of given species. This leads in
  situation where closely related sequences of
  given set gets relatively high weight in the
  solution.

16
Substring Parsimony Problem
Your turn Torsten...
17
Results
18
The Footprinter

• Available for free at http//bio.cs.wshington.edu
  /software.html
• Uses substring parsimony method to define
  possible motifs
• Is under constant development
• Example data at
• http//www.soe.ucsc.edu/blanchem/gh1/some_gh1.fa.
  main.html

19
Algorithm performance

• FootPrinter program
• Data set c-myc proto-oncogene upstream sequences
• k12
• d3
• n10
• L varies between 450 to 900 nucleotides
• Solution 3 distinct conserved substrings
• Computer P3 550MHz, 512 RAM

20
Algorithm performance

21
FP, Example results 1

• Metallothionein Gene family
• Promoter sequences available for wide variety of
  species
• REs have been experimentally determined in
  several species
• 4 major isoforms
• FootPrinter
• 590 bp upstream
• K7,8,9,10
• 12 highly conserved regions of wich 4 have been
  confirmed
• REs found present in most of the isoforms

22
(No Transcript)
23
FP, Example results 2

• Insulin gene family
• two rodents and a pig (two gene copies each)
• motifs with 0 mutations K8
• motifs with 1 mutation K9,10
• Footprinter
• Find 4 verified motifs
• Many were missed cause they contained too many
  mutations
• Search for motifs lengths of 12 and 15 did not
  fix the results

24
FP, insulin gene family

• Why many known binding sites were lost?
• Five categories
• No matches in other species
• Concerved regions are shorter than looked for
• Insertions and deletions not allowed
• Motifs do not meet statistical thresholds
• Motifs with internal variable sequence

25
Other computational methods

• Clustalw
• Tree based global multiple alignement tool
• Good results with closely related sequences
• Bad results if sequences are diverged
• As fast as Footprinter on this test set
• Diaglin
• Segment based multiple alignment tool
• Good results cause of search for short concerved
  regions
• Same set of found motifs as footprinter on this
  test set
• 10 times slower than Footprinter on large
  datasets

26
Other computational methods

• MEME
• Motif finding tool
• Searches for motifs with high information content
• Motifs my appear in different order in sequences
• Approximately same set of found motifs as
  Footprinter
• As fast as Footprinter on this test set

27
Future work

• Lack of sequences
• Inaccuracies in phylogenetic tree
• More need to know on what to expect
• More filtering accuracy
• How unusually well concerved the region is
• Predicting pairs and triplets

28
Questions