1. Lecture #7: FASTA & LFASTA
BIOINF 2051
Fall 2002
2/22/2019

2. Dot Plot
Alpha chain
vs.
Beta chain of
Human
Hemoglobin

3. FASTA and LFASTA Pearson and Lipman (1988)
FASTA – program that calculates the initial and optimal similarity scores between two sequences
LFASTA – program for detecting local similarities – finds multiple alignments between smaller portions of two sequences

4. The FASTA algorithm Four steps: Identify regions of similarity:
Using the ktup parameter which specifies # consecutive identities required in a match
10 best diagonal regions found based on #matches and distance between matches
Rescore regions and identify best initial regions
PAM250 or other scoring matrix used for rescoring the 10 diagonal regions identified in step 1 to allow for conservative replacements and runs of identities shorter than ktup
For each the best diagonal regions, identify “initial region” that is best scoring subregion

5. The FASTA algorithm Optimally join initial regions with scores > T
Given: location of initial regions, scores, gap penalty
Calculate an optimal alignment of initial regions as a combination of compatible regions with maximal score
Use resulting score to rank the library sequences
Selectivity degradation limited by using initial regions that score greater than some threshold T
Align the highest scoring library sequences using modification of global and local alignment algorithms
Considers all possible alignments of the query and library sequence that falls within a band centered around the highest scoring initial region

6. LFASTA
FASTA – reports only one highest scoring alignment between two sequences
LFASTA – local sequence comparison tool that can identify multiple local alignments between 2 sequences
Optimal algorithms for sensitive local sequence comparison are computationally intensive in terms of time and memory

7. LFASTA vs. FASTA
LFASTA uses same first 2 steps for finding initial regions as FASTA, except:
Instead of saving 10 initial regions, LFASTA saves all diagonal regions with similarity scores > some threshold
Construction of optimized alignments
Instead of focusing on a single region, LFASTA computes a local alignment for each initial region
Also, apart from band around initial region, LFASTA considers potential sequence alignments for some distance before and after the initial region.

8. Self-comparison of myosin heavy chain from C. elegans
See plot from a local similarity self-comparison of the myosin heavy chain (NBRF code MWKW) using the PAM 250 matrix
The amino-terminal half of the molecule forms a large globular head without any periodic structure
The symmetrical parallel lines along the C-terminal half correspond to the 28-residue repeat responsible for the a-helical coiled-coil structure of the rod segment