Professional Documents
Culture Documents
Presentationforblast Algorithm
Presentationforblast Algorithm
NAME
A.K.M.ASADUZZAMAN
KOUSHIK ROY
MD.ZAHID HASAN
MD.ASIF-AL-FAHAD
BLAST
Introduction
Suppose you have acquired a DNA/Protein sequence derived
from a sample of some environments such as lake, pond or plant.
Sequencing process
KLMNTRARLIVHISG
LTRK…………………………
……………………
Your sequence
Cell Samples
Introduction
• Or you might get a DNA/Protein sequence from a database such as
NCBI/EMBL/Swiss-Prot. You might also find an interesting
gene/sequence from a journal.
KLMNTRARLIVHISG
LTRK…………………………
……………………
Your sequence
Introduction
• In that case, you might want to know if the sequence that you have,
already exists or is similar to some sequences in a database, may be
down to a particular organism database.
Your sequence
Already in Similar?
here?
1. Local Alignment
2. Global Alignment
Your Sequence Unknown Sequence
KLMNTRARLIVHISGLTRK
????????????????????????????
Introducing BLAST (Basic Local
Alignment Search Tool)
BLAST tool is used to compare a query sequence with a
library or database of sequences.
It uses a heuristic search algorithm based on statistical
methods. The algorithm was invented by Stephen
Altschul and his co-workers in 1990.
BLAST programs were designed for fast database
searching.
BLAST Algorithm
BLAST Algorithm
BLAST Algorithm
BLAST Algorithm (Protein)
Query Sequence
Length 11 Y A N C L E H K M G S
L E H
E H K
W= 3
Y A N H K M
H K M
H K M
This generates 11 – 3 + 1 = 9 words
BLAST Algorithm Example
For each word from a window = 3, generate neighborhood words using
BLOSUM62 matrix with score threshold = 11
3 Amino Acids
L E H 20 x 20 x 20 203 alignments
17 L E H
13 L K H
Score 12 C E H
threshold 11 QE H All aligned with
(cut off here) LEH using
10 L MH
BLOSUM62
9 L F H (then sorted by
scores)
9 L E R
9 D EH
...
Sorted by scores
BLAST Algorithm Example
L E H Database sequences
L E H
Word List L E H
L K H
L E H L K H
L K H L E H C E H
L E H C E H
C E H
QEH
QE H
DAPCQEHKRGWPNDC
Exact matches of words from the word list to the database sequences
For each exact word match, alignment is extended in both directions
to find high score segments. Query = Y A N C L E H K M G S
Extended in the right direction Max drop off score X= 2
Q E H K MG S
D A P C Q E H K R G W P N D C
5 5 8 5 -1 6 -3
5 10 18 23 22 28 25
30
Score drop = 3 > X
Accumulated Score
25
Score drop = 1 <= X
20
Trim to max
15
10
5
0
Q-Q E-E H-H K-K M-R G-G S-W
For each exact word match, alignment is extended in both directions
to find high score segments. Query = Y A N C L E H K M G S
Extended in the left direction Max drop off score X= 2
Y A N CQ E H K MG
D A P C Q E H K R G W P N D C
-3 4 -2 9 5 5 8
26 2925 27 18 13 8
35
Accumulated Score
30
Score drop = 3 > X
Score drop = 2 <= X
25
20
15
10
5
0
H-H E-E Q-Q C-C N-P A-A Y-D
BLAST Algorithm Example
Maximal Segment Pair (MSP)
A N C Q E H K M G
A P C Q E H K R G
4 -2 9 5 5 8 5 -1 6
Sorted by alignment
Each match has its own E-Value
scores
BLAST Algorithm
Expect Value (E-Value)
E-Value: The number of MSPs with similar score or
higher that one can EXPECT to see by chance alone
when searching a database of a particular size.
BLAST Algorithm
Expect Value (E-Value)
For example: if the E-Value is equal to 10 for a
particular MSP with score S, one can say that
actually…about 10 MSPs with score >= S can just
happen by chance alone (for any query sequence).
So most likely that our MSP is not a significant match
at all.
BLAST Algorithm
Expect Value (E-Value)
If E-Value if very small e.x. 10-4 (very high score S), one
can say that it is almost impossible that there would be
any MSP with score >= S.
Thus, our MSP is a pretty significant match
(homologous).
BLAST Algorithm
E-Value Calculation
First: Calculate bit score
= Bit Score
m = query length
n = length of database
BLAST Algorithm
E-Value Interpretation
• E-values of 10-4 and lower indicate a significant
homology.
Blastp 3
Blastn 11
Blastx 3
Tblastn 3
Tblastx 3
BLAST Suggestion
Where possible use translated sequence (Protein).
Split large query sequence (if > 1000 for DNA, >200 for
protein) into small ones.
If the query has low complexity regions or repeated
segments, remove them and repeat the search.
IVLKVALRPVLRPVLRPVWQARNGS