Gapped BLAST and PSI-BLAST: a new generation of protein database search programs - PowerPoint PPT Presentation

About This Presentation

Title:

Gapped BLAST and PSI-BLAST: a new generation of protein database search programs

Description:

Gapped BLAST and PSI-BLAST a new generation of protein database search programs Presented by Outline BLAST 1.0 BLAST 2.0 ... – PowerPoint PPT presentation

Number of Views:162

Avg rating:3.0/5.0

Slides: 87

Provided by: musashi

Category:

more less

Transcript and Presenter's Notes

Title: Gapped BLAST and PSI-BLAST: a new generation of protein database search programs

1
Gapped BLAST and PSI-BLASTa new generation of
protein database search programs

Presented by
???
???
???
???

2
Outline

BLAST 1.0
BLAST 2.0
The two-hit method
Gapped alignment
PSI-BLAST
Performance evaluation
Discussion and Conclusion
NCBI website

3
Statistical preliminaries

HSP High-scoring segment pair
Locally optimal pair
S (?S - ?K) / ?2
S normalized score
Pi background probability that amino acids
occur randomly at all position
sij score for aligning each pair of amino acids
I and j
K minor constant
? constant to adjust for matrix
sij and Pi ? K and ?

E N / 2S
E number of distinct HSPs with normalized score
at least S
N mn is search space
S log2(N/E)
qij PiPje?uSij
qij target frequency of aligned pair of letters
(i, j) with HSP, high-scoring segment paris
?u the unique positive number

5
BLAST

Basic Local Alignment Search Tool(by Altschul,
Gish, Miller, Myers and Lipman)
The BLAST program are widely used tools for
searching protein and DNA database for sequence
similarities
BLAST is a heuristic that attempts to optimize a
specific similarity measure.
The central idea of the BLAST algorithm is that a
statistically significant alignment is likely to
contain a high-scoring pair of aligned words.

6
The maximal segment pair measure

MSP(maximal segment pair) the highest scoring
pair of identical length segments chosen from 2
sequences
for DNA Identities 5 Mismatches -4
for protein BLOSUM62
BLAST heuristically attempts to calculate the MSP
score.
DP is O(mn) ,but BLAST is O(m)

the highest scoring pair
7
BLAST 1.0

Build the hash table for Sequence A.
Scan Sequence B for hits.
Extend hits.

8
Step 1 Build the hash table for Sequence A.
(3-tuple example)

For DNA
Seq. A ACGTAGTA
12345678
AAA
AAC
..
ACG 1
..
AGT 5
..
CGT 2
..
GTA 3 6
..
TAG 4
..
TTT

For protein
Seq. A YGGFMAdd xyz to the hash table if
Score(xyz, YGG) ? TAdd xyz to the hash table
if Score(xyz, GGF) ? TAdd xyz to the hash
table if Score(xyz, GFM) ? T
T threshold parameter
High T yelds greater speed,
but weak similarities

Hash table
9
List all words in query
YGGFMTSEKSQTPLVTLFKNAIIKNAHKKGQ YGG GGF GFM
FMT MTS TSE SEK
10
Augment word list
YGGFMTSEKSQTPLVTLFKNAIIKNAHKKGQ YGG GGF GFM
FMT MTS TSE SEK
AAA AAB AAC YYY
11
G G F A A A 0 0 -2 -2
Non-match
BLOSUM62 scores
G G F G G Y 6 6 3 15
Match
A user-specified threshold determines which
three-letter words are considered matches and
non-matches.
12
YGGFMTSEKSQTPLVTLFKNAIIKNAHKKGQ YGG GGF GFM
FMT MTS TSE SEK
GGI GGL GGM GGF GGW GGY
13
Store words in search tree
Augmented list of query words
Does this query contain GGF?
Search tree
O(1) time
Yes, at position 2.
14
Search tree
G
G
L
M
F
W
Y
15
Scan the database
x
x
x
x
Query sequence
x
x
x
x
Database sequence
16
Extend hit
L P P Q G L L Query sequence M P P E G L L
Database sequence ltwordgt 7 2 6
BLOSUM62 scores word score
15 lt--- ---gt 2 7 7 2 6 4 4 HSP SCORE 32
hit
Extend
This is done by extending a hit in both
directions, until the running alignments score
has dropped more than Xbelow
17
BLAST 2.0The two-hit method

BLAST 1.0
Extension step typically accounts for gt90 of
BLAST execution time
Observations
A HSP of interest is much longer than a single
word pair
Entail multiple hits on the same diagonal and
within short distance of one another
Invoke an extension only when two non-overlapping
hits are found within distance A on the same
diagonal

Recenti the most recent hit found on the ith
diagonal (always increasing)

overlap
19

T must to be lowered
one-hits W3 ,T13
Two-hit W3 ,T11
More one-hits while the majority are dismissed
Sensitivity
For HSPs with at least 33 bits, the two-hit
heuristic is more sensitive
Speed(two-hit)
Generates on average 3.2 times as many hit, but
only 0.14 times as many hit extension(decide
whether a hit need be extended)
Twice as rapid as one-hit

20
Gapped alignment

Original BLAST find several distinct HSPs
All HSPs related to one alignment should be found
Gapped BLAST tolerate a much higher chance of
missing any single moderately scoring HSP
Seeking a single gapped alignment, rather than a
collection of umgapped ones
For example, result should gt 0.95, p miss prob
of HSP
Orignial with 2 HSP (1-p)(1-p)gt0.95? plt0.025
Now p2lt0.05?p0.22
T can be raised ? faster
Now
Find one HSP only seed, than use 2-hit

21
Gapped alignment (contd)

A gapped extension takes much longer to execute
than an ungapped extension, but by performing
very few of them the fraction of the total time
could be kept low.
Trigger a gapped extension for any HSP exceeding
score Sg
Sg should be set at 22 bits (150)

22
Original BLAST locates only the first and the
last ungapped aligment, E-value gt 50 times
23
Gapped Local Alignments
http//binfo.ym.edu.tw/post/internet/gap_blast.htm

24
actaactattacagactaactattacagactaactataca actaactat
tacggactaacttacagactaactaaaca
Before Gap Insertion actaactattacagactaactattacaga
ctaactataca
actaactattacggactaacttacagactaactaaaca Percent
Identity 24/40 0.6
After Gap Insertion actaactattacagactaactattacagac
taactataca
actaactattacggactaact--tacagactaactaaaca Perce
nt Identity 36/40 0.9
25
Gapped Local Alignments

Start from a single aligned pair of residues,
called the seed.

26
Gapped expansion

Find out ungapped region with highest alignment
score.
If the length of the ungapped region larger than
Sg, then try using DP
Use its central residue pair as the seed.
Gapped extension is invoked less than once per 50
database sequences.

27
PSSM
28

conserved regions
same protein family
some regions are very similar
the structure and functionality typical to this
family

29
PSI-BLAST (Position-Specific Iterated BLAST)
1 Select a query and search it against a
protein database 2 PSI-BLAST constructs a
multiple sequence alignment then creates a
profile or specialized position-specific scoring
matrix (PSSM) 3 The PSSM is used as a query
against the database 4 PSI-BLAST estimates
statistical significance (E values) 5 Repeat
steps 3 and 4 iteratively, typically 5
times. At each new search, a new profile is used
as the query.
PSSM
PSSM
From http//bioweb.pasteur.fr/seqanal/blast/intro
-uk.html
30
Score matrix architecture

Each matrix has length precisely equal to that of
the original query sequence.

31
Multiple alignment construction

E-value lt 0.01 from the output of BLAST output.
Any row identical to the query segment with which
it aligns is purged.
Only one copy is retained of any rows that are
above 98 identical to one another.

32
Multiple alignment construction

Pairwise alignment columns that involve gap
characters inserted into the query are simply
ignored.
So M has exactly the same length as the query.

33
Multiple alignment construction

The matrix scores for a given alignment column
should depand not only upon the residues
appearing there.
The set R of sequences it includes to be exactly
those that contribute a residue to column C.
The columns of MC to be just those columns of M
in which all the sequences of R are represented.

34
(No Transcript)
35
Sequence weights

A large set of closely related sequences carries
little more information than a single member, but
its size may allow it outvote a small number of
more divergent sequences.
One way is to assign weights.
Gap characters are treated as a 21st distinct
char.

36
Sequence weights

In constructing matrix scores, not only a
columns observed residue frequencies are
important.
Estimate the relative number NC of independent
observations constituted by the alignment MC.
NC the mean number of different residue types.

a large number of independent sequences, the
estimate of Qi should converge simply to the
observed frequency of residue i in that column.
Pseudocount frequencies
Estimate Qi by

38
Performance Evaluation
39
(No Transcript)
40
Gapped BLAST 1. 3X faster than original BLAST,
finds more 2. gt100X
faster than S-W, misses only 8, same
scores PSI-BLAST 1. faster than original BLAST,
40X faster than S-W, much more
sensitive 2. multiple
iterations is even better, better for
non-redundant database of NCBI
3. slower than gapped BLAST time
for construction of PSSM
41
PSI-BLAST Examples(1)
??????????,??HIT??query, a BLAST
1.
search of SWISS-PROT reveals hits with Elt0.01
only to other HIT proteins.
2. A PSI-BLAST search, using PSSM generated by
yields the E-value of 2X10-4 for
uridylyltransferase.
42
PSI-BLAST Examples(2)BRCT proteins
43
(No Transcript)
44
Seven recent additions to the protein databases
as members of BRCT superfamily
45
Discussion
46
Possible future improvement Gap costs

Allows a gap to involve residues in both
sequences rather than just one
A gap in which k residues are inserted or deleted
and j pairs of residues are left unaligned
receives the score (abkcj)

47
Possible future improvementRealignment

??????threshold?pairwise alignment?????multiple
alignment,?????the most significant??initial
multiple alignment and PSSM,?????rescore and
realign database sequences that received lower
scores
??
Improve weaker pairwise alignments
False positive can be downgraded by an improved
matrix
False negative can have their scores increased

48
Conclusion

Gapped version of BLAST is faster than original
one, and able to produce gapped alignments.
PSI-BLAST greatly increase sensitivity to weak
but biologically relevant sequence relationships.
PSI-BLAST retains the ability to report accurate
statistics, per iteration runs in times not much
greater than gapped BLAST, and can be used both
iteratively and fully automatically.

49
NCBI

Books
Pudmed
Blast
(1)Nucleotide
-- Quickly search for highly similar sequences
-- Nucleotide-nucleotide BLAST
(2)Protein
-- Protein-protein BLAST
(3)Translated
-- Translated query vs. Protein database
(4)Special
-- Align two sequences

50
(No Transcript)
51
(No Transcript)
52
NCBI

Books
Pudmed
Blast
(1)Nucleotide
-- Quickly search for highly similar sequences
-- Nucleotide-nucleotide BLAST
(2)Protein
-- Protein-protein BLAST
(3)Translated
-- Translated query vs. Protein database
(4)Special
-- Align two sequences

53
(No Transcript)
54
(No Transcript)
55
(No Transcript)
56
NCBI

Books
Pudmed
Blast
(1)Nucleotide
-- Quickly search for highly similar sequences
-- Nucleotide-nucleotide BLAST
(2)Protein
-- Protein-protein BLAST
(3)Translated
-- Translated query vs. Protein database
(4)Special
-- Align two sequences

57
Database searching
Sequence database
Query
Targets ranked by score
Sequence comparison algorithm
58
(No Transcript)
59
(No Transcript)
60
(No Transcript)
61
(No Transcript)
62
(No Transcript)
63
NCBI