Computational Biology Protein Threading

1
Computational BiologyProtein Threading
Instructor Prof. Jesús A. Izaguirre Textbook
Introduction to Computational Molecular
BiologySetubal and Meidanis, Chapter
8.3 References R. H. Lathrop and T. F. Smith,
Global Optimum Threading with Gapped Alignment
and Empirical Pair Score Functions, J. Mol.
Biol. (1996) 255, 641-665 Z. Luthey-Schulten,
Bioinformatics, NSF Summer School 2003,
http//www.ks.uiuc.edu/Training
2
Motivation

• Study sequence-structure alignment
• Proteins with similar sequence of aminoacids
  often have similar structure
• However, some structurally and functionally
  similar proteins have vastly different structure

3
(No Transcript)
4
(No Transcript)
5
(No Transcript)
6
Mathematical problem

• This problem is NP-hard when variable gaps and
  amino acid interaction energy functions are
  allowed
• Lathrop Smith (1996) used a Branch and Bound
  technique for solving the problem

7
Building the Scaffold I

• (A) Two structurally similar proteins and a
  common core of four secondary structure segments
  (dark lines I-L)

8
Building the Scaffold II

• (B) Abstract model showing the interactions
  (spatial adjacencies)

9
One possible threading with novel sequence

• (C) Small circles represent amino acids holes
  in the abstract scaffold. The t are the
  indices of the target residues placed into the
  first element of the core (i-l)

10
Defining and splitting sets of threadings
(branching)

• (D) Sets of threadings are defined by lower and
  upper bounds b,d for the first element of the
  sequence amino acid residue. A set is split by
  choosing a core segment (here, I), a splitting
  point (dark interior arrow) and subintervals (lt,
  , gt)

11
Illustration of branch and bound technique

• Illustration of the splitting of search space
  into subsets
• Numbers closest to the set indicate the current
  lower bound
• List of current subsets kept on priority queue or
  heap ordered from lowest bound

12
Threading Algorithm I

• Initialization
• Compute a lower bound for the set of all
  threadings
• Initialize priority queue to contain one entry,
  the set of all threadings with its lower bound

13
Threading Algorithm II

• Iteration
• Get set from priority queue having lowest bound
• If set contains one threading, found solution
• Else, split set into smaller subsets, compute
  lower bound for these, and insert into priority
  queue

14
Performance of the B B threading algorithm

• Log_10 of search time vs. Log_10 of search space
  size for 60 proteins from the PDB database (ref.
  Lathrop Smith, 96, Fig. 2)
• The line marks a 2 hour computational limit

15
Experiment to motivate need for threading

• Compare sequence vs. sequence alignment and
  sequence vs. structure alignment (threading) for
  two structurally related proteins that have not
  much sequence similarity

16
Bioinformatics Tutorial I

• We will compare the sequences and structures of
  two proteins.
• Go to the PDB database in two separate windows
  http//www.rcsb.org/pdb/
• In one window, search for PDB ID 1sbp and in the
  other 1pot.
• In the Sulfate Binding Protein window, click
  Sequence Details on the left then Download in
  FASTA Format.

17
Bioinformatics Tutorial II

• In a new window, go to the BLAST page
  http//www.ncbi.nlm.nih.gov/blast/
• Click Standard protein-protein BLAST blastp.
• Paste in the sequence for 1sbp.
• Select pdb from the database menu.

18
Bioinformatics Tutorial III

• Deselect Do CD-Search.
• Set an expect threshold of 1000.
• In the Format section of the options, specify
  1000 descriptions. It is important to set this in
  the Format section of the first page of BLAST
  options, not in the subsequent page that appears.
  
• Now click BLAST then Format.

19
Bioinformatics Tutorial IV

• Approximately at what ranking in the results did
  1pot appear? You can search for 1pot on the web
  page of results using the browser's Find command.
  
• What is the ratio between the score for 1pot and
  the maximum possible score for a result in this
  BLAST search?
• We will now perform threading of the 1sbp
  sequence. Go to the PredictProtein page at
  http//cubic.bioc.columbia.edu/predictprotein/pred
  ictprotein.html.

20
Discussion

• At what rank does 1pot appear in the results?
  Just after the start of the AGAPE section, you
  will find a useful explanation explanation of the
  columns in the table.
• What is the ratio between the score for 1pot and
  the maximum possible score for a result in the
  threading search?
• How can you explain the difference between the
  rankings and score ratios seen in the BLAST and
  threading search?