1

Introduction to BLAST

Utah State University – Fall 2017

Statistical Bioinformatics (Biomedical Big Data)

Notes 12

2

References

Chapter 2 of Biological Sequence Analysis

(Durbin et al., 2001)

http://blast.ncbi.nlm.nih.gov/Blast.cgi

3

Basic Local Alignment Search Tool

Finds regions of similarity between a query

sequence and a growing database

Breaks query & database into fragments:

(words)

Starts out by aligning fragments, then

extends alignment

Not optimal alignment, but a good heuristic

approachsimplified, rule of thumb

4

Recall FASTA file from pairseqsim library

library(Biostrings)

f1 <- "C://folder//ex.fasta" # see slide 20 of Notes 11

ff <- readAAStringSet(f1, "fasta")

as.character(ff[1])

At1g01010 NAC domain protein, putative

MEDQVGFGFRPNDEELVGHYLRNKIEGNTSRDVEVAISEVNICSYDPWNLRFQSKYKSRDA

MWYFFSRRENNKGNRQSRTTVSGKWKLTGESVEVKDQWGFCSEGFRGKIGHKRVLVFLDGR

YPDKTKSDWVIHEFHYDLLPEHQRTYVICRLEYKGDDADILSAYAIDPTPAFVPNMTSSAG

SVVNQSRQRNSGSYNTYSEYDSANHGQQFNENSNIMQQQPLQGSFNPLLEYDFANHGGQWL

SDYIDLQQQVPYLAPYENESEMIWKHVIEENFEFLVDERTSMQQHYSDHRPKKPVSGVLPD

DSSDTETGSMIFEDTSSSTDSVGSSDEPGHTRIDDIPSLNIIEPLHNYKAQEQPKQQSKEK

VISSQKSECEWKMAEDSIKIPPSTNTVKQSWIVLENAQWNYLKNMIIGVLLFISVISWIIL

VG

5

We’ll start here, with

protein BLAST:

(blastp)

http://blast.ncbi.nlm.nih.gov/Blast.cgi

6

paste sequence

here

(or accession /

gi number)

nr database is

most general

click here to start

search

for querying

specific

subsequences

set parameters (next slide)

7

select scoring

scheme here

# “significant”

matches

expected just

by chance in

database

fragment sizes to

begin alignments

8

Wait for Results

Large database to search- time also: “traffic”-dependent

This can be automated somewhat(write scripts to deliver & format results)

Can arrange for e-mail to link of final results

9

Visualize

alignments,

colored by

score

Scroll down

or click on

alignments

to see more

information

most

interested in

top scoring

alignments

10

We already knew

this one; probably

more interested in

other “good”

alignments

11

Normalized alignment score

The expected # of sequences to score higher than this one in the

database search, just by chance

links to

gene

info

…

12

Another “good” alignment

Distribution of maximum score

13

Think of alignment score as sum of indep. random

variables.

Then the distribution of this score can be considered

approximately normal. (why?)

The asymptotic distribution of the maximum MN of a

series of N independent normal random variables is

the extreme value distribution:

Use this to calculate the probability that the best

match from a search of N unrelated sequences has

score greater than our observed largest score S.

x

N eNKMP exp

14

E-value of a local, ungapped alignment

system scoringfor scale natural

size space searchfor scale natural

residues) in database, of length(or sequence database of length

sequencequery of length

where

,exp

: theis Sleast at of

scores withs HSP'ofnumber expected theThen

).alignments local ungapped scoring- top theof (one

pair"segment scoring-high"a be Let HSP

K

n

m

SnmKE

E-value

depend on qa and s(a,b)

2log

log

:is scorebit theaside,an As

KSS

15

Statistical Significance of an Alignment

E

E

a

eXPXP

eE

EXP

a

EEaXP

EXEEPoissonX

X

1010

!0)exp(0

!)exp(

][ );(~

Sleast at score withs HSP'of #

0

Probability of observing a result at least as extreme as what was

seen, just by chance, when the sequences are all unrelated (Null)

= P-value

Still need to look at biological similarity

16

Estimated values for λ & K

H = relative “entropy” of target

and background residue

frequencies

(like the average information

available per position to

distinguish an alignment from

chance)

17

18

Sub-tree

19

Other resources to consider

Altschul et al. (1990) “Basic Local Alignment

Search Tool.” Journal of Molecular Biology

215:403-410.

Mitrophanov & Borodovsky (2006) “Statistical

significance in biological sequence analysis.”

Briefings in Bioinformatics 7(1):2-24

20

Nexus file format

21

(NOTE: can’t have duplicate tip labels; added (1), (2), & (3);

http://www.stat.usu.edu/jrstevens/bioinf/tree1.nex )

Phylogenetic trees in R

22

23

library(phylobase)

tree <- readNexus("C:\\folder\\tree1.nex")

treePlot(tree, type="phylogram")

(We may return to these trees in another discussion)

24

Summary

BLAST

look at finding global alignment

from smaller, local alignments

E-value

expected number of better scores

just by chance,

when sequences are not related

Statistical significance ≠ biological relevance