Sequence database searching

1
Sequence database searching Homology searching

• Dynamic Programming (DP) too slow for repeated
  database searches.
• Therefore fast heuristic methods
• FASTA
• BLAST and PSI-BLAST
• QUEST
• HMMER
• SAM-T98

Fast heuristics
Hidden Markov modelling
2
FASTA

• Compares a given query sequence with a library of
  sequences and calculates for each pair the
  highest scoring local alignment
• Speed is obtained by delaying application of the
  dynamic programming technique to the moment where
  the most similar segments are already identified
  by faster and less sensitive techniques
• FASTA routine operates in four steps

3
FASTA

• Operates in four steps
• Rapid searches for identical words of a user
  specified length occurring in query and database
  sequence(s) (Wilbur and Lipman, 1983, 1984). For
  each target sequence the 10 regions with the
  highest density of ungapped common words are
  determined.
• These 10 regions are rescored using Dayhoff
  PAM-250 residue exchange matrix (Dayhoff et al.,
  1983) and the best scoring region of the 10 is
  reported under init1 in the FASTA output.
• Regions scoring higher than a threshold value and
  being sufficiently near each other in the
  sequence are joined, now allowing gaps. The
  highest score of these new fragments can be found
  under initn in the FASTA output.
• full dynamic programming alignment (Chao et al.,
  1992) over the final region which is widened by
  32 residues at either side, of which the score is
  written under opt in the FASTA output.

4
FASTA output example
DE METAL RESISTANCE PROTEIN YCF1 (YEAST CADMIUM
FACTOR 1). . . . SCORES Init1 161 Initn 161
Opt 162 z-score 229.5 E() 3.4e-06
Smith-Waterman score 162 35.1 identity in 57
aa overlap
10 20 30 test.seq
MQRSPLEKASVVSKLFFSW
TRPILRKGYRQRLE


YCFI_YEAST CASILLLEALPKKPLMPHQHIHQTLTRRKPNPY
DSANIFSRITFSWMSGLMKTGYEKYLV 180
190 200 210 220 230
40 50 60
test.seq LSDIYQIPSVDSADNLSEKLEREWDRE

YCFI_YEAST EADLYKLPRNFSSEELSQKLEKNWENELKQKSN
PSLSWAICRTFGSKMLLAAFFKAIHDV 240
250 260 270 280 290

5
FASTA

• (1) Rapid identical word searches
• Searching for k-tuples of a certain size within a
  specified bandwidth along search matrix
  diagonals.
• For not-too-distant sequences (gt 35 residue
  identity), little sensitivity is lost while speed
  is greatly increased.
• Technique employed is known as hash coding or
  hashing a lookup table is constructed for all
  words in the query sequence, which is then used
  to compare all encountered words in each database
  sequence.

6
FASTA

• The k-tuple length is user-defined and is usually
  1 or 2 for protein sequences (i.e. either the
  positions of each of the individual 20 amino
  acids or the positions of each of the 400
  possible dipeptides are located).
• For nucleic acid sequences, the k-tuple is 5-20,
  and should be longer because short k-tuples are
  much more common due to the 4 letter alphabet of
  nucleic acids. The larger the k-tuple chosen, the
  more rapid but less thorough, a database search.

7
BLAST

• blastp compares an amino acid query sequence
  against a protein sequence database
• blastn compares a nucleotide query sequence
  against a nucleotide sequence database
• blastx compares the six-frame conceptual protein
  translation products of a nucleotide query
  sequence against a protein sequence database
• tblastn compares a protein query sequence against
  a nucleotide sequence database translated in six
  reading frames
• tblastx compares the six-frame translations of a
  nucleotide query sequence against the six-frame
  translations of a nucleotide sequence database.

8
BLAST

• Generates all tripeptides from a query sequence
  and for each of those the derivation of a table
  of similar tripeptides number is only fraction
  of total number possible.
• Quickly scans a database of protein sequences for
  ungapped regions showing high similarity, which
  are called high-scoring segment pairs (HSP),
  using the tables of similar peptides. The initial
  search is done for a word of length W that scores
  at least the threshold value T when compared to
  the query using a substitution matrix.
• Word hits are then extended in either direction
  in an attempt to generate an alignment with a
  score exceeding the threshold of S, and as far as
  the cumulative alignment score can be increased.

9
BLAST

• Extension of the word hits in each direction are
  halted
• when the cumulative alignment score falls off by
  the quantity X from its maximum achieved value
• the cumulative score goes to zero or below due to
  the accumulation of one or more negative-scoring
  residue alignments
• upon reaching the end of either sequence
• The T parameter is the most important for the
  speed and sensitivity of the search resulting in
  the high-scoring segment pairs
• A Maximal-scoring Segment Pair (MSP) is defined
  as the highest scoring of all possible segment
  pairs produced from two sequences.

10
PSI-BLAST

• Query sequences are first scanned for the
  presence of so-called low-complexity regions
  (Wooton and Federhen, 1996), i.e. regions with a
  biased composition likely to lead to spurious
  hits are excluded from alignment.
• The program then initially operates on a single
  query sequence by performing a gapped BLAST
  search
• Then, the program takes significant local
  alignments found, constructs a multiple alignment
  and abstracts a position specific scoring matrix
  (PSSM) from this alignment.
• Rescan the database in a subsequent round to find
  more homologous sequences Iteration continues
  until user decides to stop or search has
  converged

11
PSI-BLAST iteration
Query sequence
Q
xxxxxxxxxxxxxxxxx
Gapped BLAST search
Query sequence
Q
xxxxxxxxxxxxxxxxx
Database hits
A C D . . Y
PSSM
Pi Px
Gapped BLAST search
A C D . . Y
PSSM
Pi Px
Database hits
12
PSI-BLAST output example
13
Multiple alignment profiles Gribskov et al.
1987A way to represent multiple alignment
consensus
i
A C D ? ? ? W Y
0.3 0.1 0 ? ? ? 0.3 0.3
Gap penalties
0.5
1.0
Position dependent gap penalties
14
Normalised sequence similarity
The p-value is defined as the probability of
seeing at least one unrelated score S greater
than or equal to a given score x in a database
search over n sequences. This probability
follows the Poisson distribution (Waterman and
Vingron, 1994)
P(x, n) 1 e-n?P(S? x), where n is the
number of sequences in the database Depending on
x and n (fixed)
15
Normalised sequence similarityStatistical
significance
The E-value is defined as the expected number of
non-homologous sequences with score greater than
or equal to a score x in a database of n
sequences E(x, n)
n?P(S ? x) if E-value 0.01, then the expected
number of random hits with score S ? x is 0.01,
which means that this E-value is expected by
chance only once in 100 independent searches over
the database. if the E-value of a hit is 5, then
five fortuitous hits with S ? x are expected
within a single database search, which renders
the hit not significant.
16
Normalised sequence similarityStatistical
significance

• Database searching is commonly performed using an
  E-value in between 0.1 and 0.001.
• Low E-values decrease the number of false
  positives in a database search, but increase the
  number of false negatives, thereby lowering the
  sensitivity of the search.

17
HMM-based homology searching

• Most widely used HMM-based profile searching
  tools currently are SAM-T98 (Karplus et al.,
  1998) and HMMER2 (Eddy, 1998)
• formal probabilistic basis and consistent theory
  behind gap and insertion scores
• HMMs good for profile searches, bad for alignment
• HMMs are slow

18
The HMM algorithms

• Questions
• What is the most likely die (predicted) sequence?
  Viterbi
• What is the probability of the observed sequence?
  Forward
• What is the probability that the 3rd state is B,
  given the observed sequence? Backward

19
HMM-based homology searching
Transition probabilities and Emission
probabilities Gapped HMMs also have insertion
and deletion states
20
Profile HMM mmatch state, I-insert state,
ddelete state go from left to right. I and m
states output amino acids d states are silent.
21
Homology-derived Secondary Structure of Proteins
(HSSP) Sander Schneider, 1991