Presentation is loading. Please wait.

Presentation is loading. Please wait.

Robert Fraser May 21, 2010. Outline Contact Maps Pattern Matching in Matrices Adaptive Analysis Experimental Results Conclusions, Future Work.

Similar presentations


Presentation on theme: "Robert Fraser May 21, 2010. Outline Contact Maps Pattern Matching in Matrices Adaptive Analysis Experimental Results Conclusions, Future Work."— Presentation transcript:

1 Robert Fraser May 21, 2010

2 Outline Contact Maps Pattern Matching in Matrices Adaptive Analysis Experimental Results Conclusions, Future Work

3 Contact Maps An abstract translational and rotational invariant representation of a proteins topology An N × N matrix C, where N is the number of amino acids in the given protein Entry C ij is a boolean, indicating whether amino acid i is in contact with amino acid j. Hippy Demo

4 Pattern Matching in Matrices Search for pattern: In large text: 11000000000000000000 11100000101000000000 01110000110000000000 00111000000001000010 10011110010011100000 00001110010001100000 11000111000000000000 10000111101000001010 10000011110000000000 0000 0010 0111 0011 0000

5 Pattern Matching in Matrices Naïve: θ(N 2 ·I·J) Uses sliding window: 11000000000000000000 11100000101000000000 01110000110000000000 00111000000001000010 10011110010011100000 00001110010001100000 11000111000000000000 10000111101000001010 10000011110000000000 0000 0010 0111 0011 0000 0000 0010 0111 0011 0000 0000 0010 0111 0011 0000 0000 0010 0111 0011 0000 0000 0010 0111 0011 0000

6 Pattern Matching in Matrices Linear: θ(N 2 +I·J ) Uses a trie: Node (i) 012345678910 f(i)00122710008 Acc(i)00001010011

7 11000000000000000000 11100000101000000000 01110000110000000000 00111000000001000010 10011110010011100000 00001110010001100000 11000111000000000000 10000111101000001010 10000011110000000000 Node (i) 012345678910 f(i)00122710008 Acc(i)00001010011

8 11000000000000000000 11100000101000000000 01110000110000000000 00111000000001000010 10011110010011100000 00001110010001100000 11000111000000000000 10000111101000001010 10000011110000000000 Expected Sublinear: O(N 2 /I +IJ) expected, O(N 2 ·I) worst case

9 Adaptive Analysis Idea came from sorting: Some sequences are easier to sort than others 12345678 56781234 13572468 45173286

10 Contact maps Shouldnt search where you wont find a relevant match: Eliminate areas along the central axis Eliminate areas that are not in secondary structures

11 Which technique is best......depends upon the characteristics of your pattern.

12 Conclusions Significant gains may be achieved by incorporating additional information into your search. Linear search performs best if you restrict the search to the interface region of a contact map. Future work: further exploiting the characteristics of the data to improve search time: i.e. The matrices are sparse!


Download ppt "Robert Fraser May 21, 2010. Outline Contact Maps Pattern Matching in Matrices Adaptive Analysis Experimental Results Conclusions, Future Work."

Similar presentations


Ads by Google