1. Indiana University, Bloomington, IN
Sequence Homology
M.M. Dalkilic, PhD
Monday, September 08, 2008
Class IV
Indiana University, Bloomington, IN
Sequence Similiarty (Computation) M.M. Dalkilic, PhD SoI Indiana University, Bloomington, IN 2008 ©

2. Outline New Programming and written homework Friday
New Reading Posted on Website
Readings [R] Chaps 5
Most Important Aspect of Bioinformatics—homology search through sequence similarity (cont’d)
Sequence Alignment
Sequence Similiarty (Computation) M.M. Dalkilic, PhD SoI Indiana University, Bloomington, IN 2008 ©

3. Introduction to Entropy
Shannon’s theory of quantifying communication
Can be derived axiomatically
Simple model
Sequence Similiarty (Computation) M.M. Dalkilic, PhD SoI Indiana University, Bloomington, IN 2008 ©

4. Introduction to Entropy
An increase in surprise means an increase in information
A decreate in surprise means a decrease in information
Since for each message set we associate a probability function
Encoding of M
Sequence Similiarty (Computation) M.M. Dalkilic, PhD SoI Indiana University, Bloomington, IN 2008 ©

5. Introduction to Entropy
An increase in surprise means an increase in information
A decrease in surprise means a decrease in information
Since for each message set we associate a probability function
Encoding of M
Sequence Similiarty (Computation) M.M. Dalkilic, PhD SoI Indiana University, Bloomington, IN 2008 ©

6. Introduction to Entropy
An increase in surprise means an increase in information
A decrease in surprise means a decrease in information
Since for each message set we associate a probability function
Sequence Similiarty (Computation) M.M. Dalkilic, PhD SoI Indiana University, Bloomington, IN 2008 ©

7. Introduction to Entropy
Can formally prove these later—not complicated.
We’ll look at multivariate entropy, conditional, and mutual information later as we examine the internals of BLAST
Sequence Similiarty (Computation) M.M. Dalkilic, PhD SoI Indiana University, Bloomington, IN 2008 ©

8. NW and SM Alignment Algorithms
Initialization Phase (the initial values of the recurrences)
Fill-in (Bottom-up recursion)
Trace-back
This reduces complexity to
Cost? We cannot guarantee the best solution—only a decent solution (at best)
This is why it is mandatory to manually inspect alignments
Sequence Similiarty (Computation) M.M. Dalkilic, PhD SoI Indiana University, Bloomington, IN 2008 ©

9. NM
Initialize top row and left column by placing the negative distance away from the start of the sequences
Fill-in
Sequence Similiarty (Computation) M.M. Dalkilic, PhD SoI Indiana University, Bloomington, IN 2008 ©

10. Sequence Similiarty (Computation) M. M
Sequence Similiarty (Computation) M.M. Dalkilic, PhD SoI Indiana University, Bloomington, IN 2008 ©

11. Sequence Similiarty (Computation) M. M
Sequence Similiarty (Computation) M.M. Dalkilic, PhD SoI Indiana University, Bloomington, IN 2008 ©

12. NM and SM Alignment
Traceback—start at right-bottom and follow arrows to left- top
finish
sequence
sequence
Start
Sequence Similiarty (Computation) M.M. Dalkilic, PhD SoI Indiana University, Bloomington, IN 2008 ©

13. Recurrence
Sequence Similiarty (Computation) M.M. Dalkilic, PhD SoI Indiana University, Bloomington, IN 2008 ©

14. SM is local alignment
Initialization of top row and left column to zeros
Cell values can only be non-negative
Traceback starts at maximum value and ends at zero
Sequence Similiarty (Computation) M.M. Dalkilic, PhD SoI Indiana University, Bloomington, IN 2008 ©

15. Affine gap scores Initial gap cost is high
Continuing gaps are constant and lower
Sequence Similiarty (Computation) M.M. Dalkilic, PhD SoI Indiana University, Bloomington, IN 2008 ©