Docstoc

Improvements in Reduce

Document Sample
Improvements in Reduce Powered By Docstoc
					Placing hydrogens by
dynamic programming
              By
         Xueyi Wang
Department of Computer Science
       UNC-Chapel Hill
    Outline
   Reduce program
   Improvements
-- Convert data structures to STL.
-- Apply dynamic programming algorithm in
  large groups.
-- Fix inconsistent results.
   Future Work
      Reduce
    J. Michael Word, Simon C. Lovell, Jane S. Richardson and
    David C. Richardson. Asparagine and glutamine: using
    hydrogen atom contacts in the choice of sidechain amide
    orientation. J. Mol. Biol. (1999) 285, 1735-1747
   Causes
-- X-ray crystal structures lack hydrogen atoms.
-- Structural ambiguity.
   Goals
-- Understand hydrophobic core packing.
-- Understand the role of hydrogen bonding in protein stability.
Movers:
Rotations of Methyl, NH3+ and Donor


   C
                            Og1

  S       Methionine
                          C

            Lysine         C
                           

   C
   
    Movers:
    Ambiguous Structure



O             N            N             O




         C                          C
                                   
     Glutamine (Asparagine, Histidine)
     Approach of Reduce
   Find the placement of hydrogen atoms that
-- maximize the number of hydrogen bonds
-- minimize the number of clashes
      Score Function in Reduce
   Maximize the function:
         4 Vol (H-bond) - 10 Vol (overlap)

-- Approximate to the van
der Waals function for an
isolated pair-wise interaction.

-- Volume is approximated
by summing dots and spikes
    Improvements

   Convert the basic data structures to
    STL structures.
   Apply dynamic programming algorithm in
    large groups (more than 2 movers).
   Fix inconsistent results caused by
    different executing orders of atoms in
    groups.
    Convert Data structures

   Original data structure is quite
    inefficient.
   STL provides a very good data
    structure for manipulating data.
-- Separate data, iterator and algorithm.
-- Easy for further improvements.
    Improvements
   Convert the basic data structures to STL
    structures.
   Apply dynamic programming
    algorithm in large groups (more than
    2 movers).
   Fix inconsistent results caused by
    different executing orders of atoms in
    groups.
    Old algorithm used by Reduce
   Partition the movers into groups that have
    no interactions between them.




   Maximize the score within each group by
    brute force.
   Several groups have > 2-3 movers.
     Dynamic Programming
   Problem:
-- proteins with large groups exist and Brute force approach
   is very inefficient.
   Solution:
-- Use the interaction graph to divide the problem and do
   dynamic programming.
   Leaver-Fay, A., Liu, Y., and Snoeyink, J., Faster
    placement of hydrogens in protein structures by
    dynamic programming, Proc. ALENEX'04, LNCS,
    Springer Verlag, 2004.
     Demo
         Og1


         angle
         170
                                   Og1   Thr 13
Thr 62    82
                           angle
                            170           angle
          39      Og1       77           176   26
         -94                21           72    -71
         -141    Ser 9      -219         57    -110
                                         42


                 From 1SBP.pdb
Interaction Graph
       Experiments

   For proteins with
    large groups,
    dynamic
    programming
    algorithm can
    significantly
    reduce the
    calculation time.
    Improvements

   Convert the basic data structures to STL
    structures.
   Apply dynamic programming algorithm in
    large groups (more than 2 movers).
   Fix inconsistent results caused by
    different executing orders of atoms
    in groups.
    Inconsistent results
   Inconsistent results can be caused by
    different execution order of atoms in two
    stages:
-- Setting initial angle stage
-- Finding optimal angle stage

   Solution: Algorithm based on dynamic
    programming.
          Inconsistent Result
   Θ1+Δθ

                                       Different
                                        calculation
Angle 1                                 sequences
          θ1                            cause
                                        different
                                        result.

    Θ1-Δθ
        Θ2-Δθ             θ2    Θ2+Δθ
                Angle 2
          New Algorithm
  Θ1+Δθ

                                      Different
                                       calculation
Angle 1                                sequences
          θ1
                                       cause the
                                       same result.


   Θ1-Δθ
       Θ2-Δθ              θ2   Θ2+Δθ
                Angle 2
    Future Work

   Apply continuous scores instead
    of contact dots for the score
    function
   Enhance efficiency
-- Rewrite and reorganize some inefficient
  subroutines.
    Continuous Score

   Calculate the
    exact overlap
    volume.
  Contact dots vs. Continuous Score




Absolute and relative errors for all pairs of radii
    The End

   Acknowledgements:
-- Dr. Michael Word in GlaxoSmithKline.
-- Andrew Leaver-Fay, Yuanxin (Leo) Liu
  and Jack Snoeyink.
-- NSF grant 0076984.

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:1
posted:5/16/2012
language:English
pages:23
fanzhongqing fanzhongqing http://
About