Genome Analysis II Comparative Genomics

1
Genome Analysis IIComparative Genomics

• Jiangbo Miao
• Apr. 25, 2002
• CISC889-02S Bioinformatics

2
Why Comparative Genomics ?

• It tells us what are common and what are unique
  between different species at the genome level.
• Genome comparison may be the surest and most
  reliable way to identify genes and predict their
  functions and interactions.
• e.g., to distinguish orthologs from
  paralogs
• The functions of human genes and other DNA
  regions can be revealed by studying their
  counterparts in lower organisms.

3
Outline

• All-against-all Self-comparison of Proteome
• Between-proteome Comparisons
• Family and Domain Analysis
• Ancient Conserved Regions (ACRs)
• Horizontal Gene Transfer
• Functional Classification of Genes
• Gene-order Comparisons

4
All-against-all Self-comparison

• How?
• Making a database of the proteome
• Use each protein as a query in a similarity
  search against the database
• (BLAST, WU-BLAST or FASTA)
• Generate a matrix of alignment scores (P or E
  value)
• A conservative cutoff E value 10e-6
• Why?
• Number of Gene Families
• This comparison distinguishes unique proteins
  from proteins arisen from gene duplication, and
  also reveals the of gene families.
• Paralogs
• Significantly matched pairs of protein sequences
  may be paralogs.

5
All-against-all Comparison Example
6
Cluster Analysis

• To sort out relationships among all of the
  proteins found to be related in the above search.
• Clustering organizes the proteins into groups by
  some objective criterion
• P or E value ( lt 0.01-0.05)
• Distance between each pair of sequences in a
  multiple seq. alignment
• ( of amino acid changes between the aligned
  seq.)
• Methods
• By Making Sub-graphs
• By Single Linkage

7
Clustering by making subgraphs

• Each protein sequence is a vertex
• Each matched pair of sequences with a significant
  score is joined by an edge
• The edges are weighted according to the P/E value
• Simple Algorithm Remove weaker links (From the
  weakest one)
• Rubin et al. (2000)
• Edges of E value gt 10-6 are removed
• Remaining subgraphs comprise sequences that share
  a significant relationship to each other but not
  to other seq.
• Criterion the group should mutually share gt 2/3
  of all of the edges from this group to all
  proteins in the proteome
• This algorithm favors the selection of proteins
  with the same domain structure reflecting that
  these proteins are most probably paralogs

8
Clustering by making subgraphs Example
9
Clustering by single linkage

• Based on the distance criterion
• A group of related sequences found in the
  all-against-all proteome comp. is subjected to a
  MSA (CLUSTALW).
• A distance matrix is made
• Use this matrix to cluster the sequence by a
  neighbor-joining algorithm
• (the same procedure as that used to make a
  phylogenetic tree)
• Cluster representation Tree or Dendrogram
• As smaller groups are chosen, the most strongly
  supported clusters are more likely to be made up
  of paralogs(?)

10
Clustering by single linkage Example
11
Core Proteome

• All-against-all comparison reveals the of
  protein/gene families in an organism.
• This number represents the core proteome of the
  organism from which all biological functions have
  diversified.

Organism of genes of gene families of duplicated genes
H. Influenzae (bacteria) 1709 1425 284
S. Cerevisiae (yeast) 6241 4383 1858
C. Elegans (worm) 18,424 9453 8971
D. Melanogaster (fly) 13,600 8065 5536
In Hemophilus, 1247 out of 1709 proteins do not
have paralogs Core proteome of the
multicellular organisms is only twice that of
yeast
12
Outline

• All-against-all Self-comparison of Proteome
• Between-proteome Comparisons
• Family and Domain Analysis
• Ancient Conserved Regions (ACRs)
• Horizontal Gene Transfer
• Functional Classification of Genes
• Gene-order Comparisons

13
Between-Proteome Comparisons Why?

• To identify orthologs, gene families, and domains
• Orthologs (proteins that share a common
  ancestry function)
• A pair of proteins in two organisms that align
  along most of their lengths with a highly
  significant alignment score.
• These proteins perform the core biological
  functions shared by the two organisms.
• Two matched sequences (X in A, Y in B) may not be
  orthologs
• (Y and Z are paralogs in B, X and Z are
  orthologs)
• Identify true orthologs
• highest-scoring match (best hit)
• E value lt 0.01
• gt 60 alignment over both proteins

14
Between-Proteome Comparisons How?

• Choose a yeast protein and perform a database
  similarity search of the worm proteome
  (WU-BLAST) a yeast-versus-worm search
• Group the worm seqs that match the yeast query
  seq with a high P value (10-10 to 10-100), also
  include the yeast query seq in the group
• From the group made in 2, choose a worm seq and
  make a search of the yeast proteome, using the
  same P limit
• Add any matching yeast seq to the group made in
  2
• Repeat 3 4 for all initially matched seqs in
  the group
• Repeat 1-5 for every yeast protein
• As 1-6, perform a comparable worm-versus-yeast
  search
• Coalesce the groups of related seqs. and remove
  any redundancies so that every sequence is
  represented only once.
• Eliminate any matched pairs in which less than
  80 of each seq is in the alignment

15
Between-Proteome Comparison Result
Cut-off P value lt 10-10 lt 10-20 lt 10-50 lt 10-100
of seq groups 1171 984 552 236
of groups with gt2 members 560 442 230 79
and of all yeast proteins (6217) represented in groups 2697(40) 1848(30) 888(14) 330(5)
and of all worm proteins represented in groups 3653(19) 2497(13) 1094(6) 370(2)
The sequences also align to 80, so they
represent highly conserved sets of genes
16
Cluster of orthologous group (COG)

• Motivation
• In the above database search, A protein seq
  will not only match the orthologous seq in the
  second proteome, but also those paralogous seqs
  of the orthologous seq.
• Objective
• To identify all matching proteins as an
  orthologous group related by both speciation
  (ortholog) and gene duplication (paralog) events.
• Meaning
• COGs usually correspond to classes of metabolic
  function
• Application (example)
• Produce a COG database by analysis of microbial
  yeast genomes
• Search a newly identified microbial protein in
  this database
• Significant match will provide an indication of
  its metabolic function

17
Comparison of Proteome to EST database

• Why?
• For many organisms(Eukaryotic), complete genome
  seq not available
• While a large collection of EST seqs are
  available
• An EST database of an organism can also be
  analyzed for the presence of gene families,
  orthologs, and paralogs.
• e.g. a protein from the yeast or fly proteome can
  be used as a query of a human EST database
• (translate EST seq in all six possible reading
  frames)
• Problem
• EST seqs are usually short( the equivalent of
  100-150 amino acids)
• Solution
• identify overlapping EST seq a longer alignment
  can be produced
• perform an exhaustive search for a protein
  family

18
Search for orthologs to a protein family in EST
database

• Retief et al. (1999) Use FAST-PAN to scan EST
  database with multiple queries from a protein
  family, sorts the alignment scores, and produces
  charts and alignments of the matches found.

• Example
• Protein family glutathione transferase proteins
• Mammalian EST database
• TFASTY3 search system
• Shown are matches of two mouse ESTs to a query
  seq

19
Search for orthologs to a protein family in EST
database

• A large number of known glutathione transferase
  proteins was first subjected to MSA, and a
  phylogenetic tree was made to identify classes of
  proteins within the family
• The object was to choose class representatives

result
Class
Flow chat
Search
20
Outline

• All-against-all Self-comparison of Proteome
• Between-proteome Comparisons
• Family and Domain Analysis
• Ancient Conserved Regions (ACRs)
• Horizontal Gene Transfer
• Functional Classification of Genes
• Gene-order Comparisons

21
Family and Domain Analysis

• What is domain?
• Proteins are modular often comprise separate
  domains
• Domains represent modules of structure and
  function
• Domain Comparison
• Comparison of the domain content of a proteome
  with that of another proteome reveals the
  biological roles of diverse domains in different
  organisms.
• Example an analysis of fly, worm, yeast
  proteomes
• 744 families and domains were common to all three
  org.
• gt 2000 fly worm proteins are multidomain
  proteins (1/3 in yeast)

22
Ancient Conserved Regions (ACRs)

• What is ACR?
• In some phylogenetically diverse groups of
  organisms, there are conserved proteins or
  protein domains that have been conserved over
  long periods of evolutionary time.
• How to find ACRs?
• Database similarity search of the SwissProt
  database with human, worm, yeast and E. coli
  genes
• Identify matches with sequence from a different
  phylum than the query sequence
• The number of ACRs may be estimated by the
  proportion of genes that match database sequence
  of known function
• e.g. 70 prokaryotic genomes contain ACRs

23
Horizontal Gene Transfer

• Horizontal Transfer (HT)
• the acquisition of genetic material from a
  different organism and these transferred material
  then becomes a permanent addition to the
  recipient
• (HT is a significant source of genome
  variation for bacteria)
• Comparisons of bacterial genomes reveal that they
  are mosaics of ancestral (vertical) and
  horizontally transferred seqs.
• 12.8 of the genome of E. coli is due to HT DNA
  (the highest level)
• How to detect HT?
• Fact each genome of bacterial species has a
  unique base composition
• HT can be detected as an island of seq with
  different composition
• If the amino acid composition of transferred
  genes is typical, these islands may be detected
  by a codon usage analysis
• The time of the transfer may be estimated by the
  degree of blend

24
Outline

• All-against-all Self-comparison of Proteome
• Between-proteome Comparisons
• Family and Domain Analysis
• Ancient Conserved Regions (ACRs)
• Horizontal Gene Transfer
• Functional Classification of Genes
• Gene-order Comparisons

25
Functional Classification of Genes

• Genes that are significantly similar in an
  organism, i.e., paralogous seqs, frequently are
  found to have a related biological function.
• Classification Scheme
• Eight related groups of E. coli genes enzymes,
  transport elements, regulators, membranes,
  structural elements, protein factors, leader
  peptides, and carriers.
• 90 of E. coli genes fell into these same
  broad categories
• Special Commission, e.g. Enzyme Commission of
  (IUBMB) provides a kind of detailed classes based
  on the biochemical reactions they catalyze
• Examine relationships among multiple enzymes that
  perform the same biochemical function in the same
  organism. (these enzymes showed variations in
  metabolic regulation of their activity)

26
Outline

• All-against-all Self-comparison of Proteome
• Between-proteome Comparisons
• Family and Domain Analysis
• Ancient Conserved Regions (ACRs)
• Horizontal Gene Transfer
• Functional Classification of Genes
• Gene-order Comparisons

27
Gene Order Comparison

• Observations about gene order
• Gene order is highly conserved in closely related
  species but becomes changed by rearrangements
  over evolutionary time
• Groups of genes that have a similar biological
  function tend to remain localized in a group or
  cluster
• Chromosomal Rearrangement
• Occasional chromosomal breaks (random chromosomal
  location)
• Random rejoining of the fragments by a DNA repair
  mechanism
• Rearrangement Analysis
• By comparing the location of orthologs

28
Chromosomal Rearrangement
29
Computational Analysis of Genome Rearrangements

• Challenges
• The number and types of rearrangements that have
  occurred
• When they occurred?
• Example a comparison of human and mouse
  chromosomes
• Computational Approach
• Genome alignment
• Alignment reduction reconstruct the number and
  types of rearrangement

30
Computational Analysis of Genome Rearrangement

• Human chromosomes were cut into gt 100 pieces and
  reassembled into a reasonable facsimile of the
  mouse chromosome.

31
Computational Analysis of Gene Rearrangement
Circular

• Lines indicate homologous position
• The more rearrangements there are, the more
  intersections will occur
• Sankoff Goldstein(1989) devised a shuffling
  model for estimating the of rearrangements
  given the of intersections.

32
Computational Analysis of Gene Rearrangement

• Assume that those rearrangements have occurred
  by some transposition or recombination events
• And identify the rearrangements by undoing
  those events.
• The goal is to minimum the number of
  rearrangements, which represents a genetic
  distance between the two genome sequences

33
Clusters of Genes on Chromosomes

• In a given organism, genes are found in a given
  order that is maintained on the chromosomes.
• On the other hand, genes with a related function
  are frequently found to be clustered at one
  chromosome location
• Example tryptophan genes in different
  prokaryotic organisms
• Observation
• At least some of the trp genes are also clustered
  together on the chromosomes of other species of
  Bacteria Archaea
• The order of genes within the cluster is
  conserved within the first four species
  (bacteria)
• The order is much less conserved in the last
  three species (Archaea)
• Gene fusions, which generate a new protein that
  performs both biochemical functions of the
  single-gene, parent proteins.

34
Clusters of Genes on Chromosomes
35
Cluster of Genes on Chromosomes

• How to identify those clusters or coordinately
  regulated genes?
• Overbeek et al. (1999)
• Perform a full reciprocal search between the
  proteomes of two org.
• Protein pairs that gave a best hit with the other
  genome had an E value lt 10-5 were identified,
  called a bidirectional best hit (BBH)
• Pairs of close BBH (PCBBH) that are within 300 bp
  of each other on the chromosomes of the
  respective organisms and that are transcribed
  from the same strand, i.e., are in a typical
  operon, were then identified
• A score for these pairs was formulated. When the
  of organisms in which the pair is observed is
  greater and the phylogenetic distance between the
  organisms is larger, this score is higher
• 40 of these pairs with higher score
  correspond to proteins that are known to act in a
  common metabolic pathway.
• ? A significant proportion of the pairs of PCBBH
  correspond to genes that have a related function
  and lie on the same pathway.