Data Searches and Sequence Alignments

1
Data Searches and Sequence Alignments

• Assessing pairwise sequence similarity BLAST
• Creation and analysis of protein multiple
  sequence alignment

2
Why do we want to perform an alignment?

• Assumption Evolutionary (phylogenetic)
  relationship
• Functional implication
• Build phylogenetic tree
• Sequences can be used for alignment
• nucleotide sequence or protein sequence

3

• Homolog A YES or NO question
• yes, share a common ancestor
• no, not related
• Similarity can be described as a fractional
  value
• AAATACGCGGTAATAGCATGCATTAGTGGT
• AATTACGCCGTAATTGCAAGCATTAGTGGT
• 26/3087 identity
• Too short to determine if they are homologous

4

• For amino acids sequence
• MATPGAGGRDKLIVASCYPVLIFIIAWQMQEP
• MHSPGAAGKERLLVASCYPVIGFILAWNSQDP
• Identity 2132 66
• Similarity -- 2832 87.5
• In general, protein sequences share over 30
  similarity are likely to be homologues.
• Usually protein sequences are longer than 50
  residues (minimal length for a domain)

5
Evaluation of two sequences

• Dot plot

6
DOROTHYCROWFOOTHODKIN DOROTHY-------HODKIN
GAP
7
Aligned local sequence
8
Tandem repeat
9
Low complexity
10
Identify exon-intron
11
Inverted repeat for terminator
12
Aligned with frame 1
frameshift
Aligned with frame 3
13
Simple alignment

• Match score and penalty

14
Gaps in alignment
15
Origination of gaps

• Insertion vs. deletion (indel) events
• One step event in evolution
• Origination penalty (open gap penalty)? higher
  penalty value
• Length penalty (gap extension penalty)? smaller
  penalty value

16
Scoring matrices -- nucleotide
Conservative substitutions are more likely to be
preserved in evolution Every mismatch should not
give the same penalty? weighted score
17
Scoring matrices amino acid residues

• PAM (point accepted mutation) matrix the scores
  are computed by the substitutions that occur in
  alignments between highly similar sequences
  (relative mutability)
• PAM unit -- the amount of evolutionary time
  required for an average of one substitution per
  100 residues to be observed
• Lower numbered PAM matrices are more suitable for
  comparing closely related proteins
• Usually PAM250 is used
• BLOSUM matrix the substitution rate are
  calculated by statistical clustering methods when
  the un-gapped related protein sequences are
  aligned
• Higher numbered BLOSUM matrices are more
  suitable for comparing closely related proteins
• Usually BLOSUM 62 is used comparing protein
  sequences with approx. 62 similarity

18
Similarity Physico-Chemical Properties of Amino
Acids
http//swift.embl-heidelberg.de/course/BSchap-4.ht
mlprin
19
(No Transcript)
20
(No Transcript)
21
Algorithms for searching the best alignment

• Exhaustive search is intractable
• Dynamic programming breaking a problem into
  sub-problems and using partial results to compute
  the final answer
• S. Needleman and C. Wunsch

22
GLOBAL AND LOCAL ALIGNMENT

• Global alignment compare two sequences entirely
• Gap penalty is assessed regardless of their
  location
• Semi-global alignment -- terminal gaps are not
  penalized

23
ACACTGATCG ACACTG----
24
GLOBAL AND LOCAL ALIGNMENT

• LOCAL ALIGNMENT
• Functional unit in sequence are more conserved
  than flexible regions
• F. Smith and M. Waterman algorithm

25

• Why do you want to do a databank search?
• Early discovery of protein purification artifacts
  
• Identification of new/unknown proteins
• Look for possible functions of an unknown protein
  
• Collect sequences for other studies
• phylogenetic analysis
• primer design
• looking for new motifs

26
Database Search
Six frame translation
27
(No Transcript)
28
Word size 4
29
Score is related to the length of match
30
Significance of the BLAST result

• E kmNe-ls

m length of query N total number of letter in
the target database l constant, dependent on
scoring matrix S score of HSP
31
Score of two random sequence at the same length
Low Score with a lot of hits? no significance
Z (score - mean)/std dev 5
Score high enough not just random hit
If the score of the alignment observed is no
better than expected from a random permutation of
the sequence, then it is likely to have arisen by
chance.
32
(No Transcript)
33
(No Transcript)
34
(No Transcript)
35
(No Transcript)
36
(No Transcript)
37
BLAST cutoffs

• Nucleotide
• Elt10-6
• Identitygt70
• Protein
• Elt10-3
• Identitygt25
• Careful around the twilight zone, use reciprocal
  BLAST or randomized sequence to BLAST

38
megablast

• Variation of blastn
• Align long or highly similar nucleotide sequences
• Locate the sequence in a contig
• Word size 28, for almost exact match
• No penalty for open gap for fast search, tend to
  get matches with more small gaps.

39
PSI-BLAST

• Position-specific-iterated BLAST
• Identify distantly related proteins
• Using position-specific score matrix (PSSM) to
  train the search, each iteration will include the
  new residue into the profile (PSSM), and use the
  new profile for next iteration (search).

40
Start from BLAST ? using high score hits to make
a PSSM ? using the PSSM for second BLAST ?
repeat for 3rd
41
Check the ones to make new PSSM