Pairwise Sequence Analysis

1
Pairwise Sequence Analysis

• Why is it useful?
• What are the underlying concepts?
• What algorithms are used?
• Which are the existing implementations?
• Limitations/Open questions

2
Why

• Given sequence A, what are its properties?
• Find any sequence(s) B similar to A, where B ?
  Sequences with known properties
• Given sequences A B, are they
• Identical?
• Similar?
• Evolutionary related?
• Given sequence A, is this a new discovery?
• Given a set of sequences, can they be clustered
  into families of related sequences?

3
Concepts

• Use of a scoring scheme to assign a value to a
  candidate alignment
• Finding the best alignment between two sequences
• Use of a probability model to assess the
  significance of the similarity

4
Concepts - Alignment

• Sequences may be aligned locally (look for
  regions/subsequences that are similar) or
  globally (align along entire length)
• The result of a local alignment may differ from
  that a global alignment

5
Concepts - Alignment

• Potentially large number of alignments to be
  considered
• Each alignment is a path through a fully
  connected dot matrix graph, i.e., a subset of the
  set of all diagonal edges, connected by
  horizontal or vertical lines. Complexity?

6
Concepts Scoring scheme

• Scheme may be based on identity or on similarity
• A match is assigned a positive score
• Gaps are assigned a penalty
• Scores are summed for every aligned position
• Highest scoring alignment(s) presented

7
Concepts Scoring scheme

• Scores are based on amino-acid substitution
  matrices (log-odds ratio of related versus random
  substitutions)
• PAM (Percent amino acid substitution) matrices
  Based on evolutionary model. E.g., PAM 1, PAM
  250.
• BLOSUM (Blocks substitution) matrices Based on
  percent identity. E.g., Blosum50, Blosum62.

8
Concepts Scoring scheme

• Identity based scoring
• Isawa --- dancer
• Isawatapdancer
• Isawa ----- danc-er
• Isawablackpanther
• Similarity based scoring
• Eyl
• dkv

9
Algorithms

• Optimal/Exact solutions
• Take longer time
• Typically used for comparing a small number of
  sequences
• (Needleman-Wunsch Smith-Waterman)
• Heuristic
• Frequently close to the optimal solution
• Rapid
• Typically used for searches large number of
  sequences
• (BLAST, FASTA)

10
Algorithms Optimal solution

• Dynamic Programming
• Generate a matrix of optimal sub-scores
• Keep a record of how sub-scores were derived
• Retrieve path from the graph with the maximum
  score by using sub-scores and the record of how
  they were derived

11
Existing Implementations

• Smith-Waterman http//www.ch.embnet.org/software/
  FDF_form.html
• SIM
• http//us.expasy.org/tools/sim-prot.html
• BLAST
• http//www.ncbi.nlm.nih.gov/BLAST/
• FASTA http//www.ebi.ac.uk/fasta33/

12
Open Questions/Limitations

• Highly similar sequences or highly dissimilar
• sequences are easily found BUT
• Gray area lies in weak similarity probability
  distribution is continuous
• Other forms of similarity (e. g., structural
  similarity) are taken into consideration to reach
  a decision
• Sequence similarity per se does not imply
  functional equivalence small difference can be
  responsible for large difference in function