BLAST

This is the Title of the Book, eMatter Edition

Copyright © 2012 O’Reilly & Associates, Inc. All rights reserved.

Where Did My Oligo Go?

|

109

So far in this chapter, we’ve just walked through most basic operations of Karlin-

Altschul statistics to provide you with the knowledge necessary to calculate bit

scores, effective lengths, and Expects. We’ve explained that BLAST uses one statisti-

cal measure to calculate the Expect of an HSP and another to calculate the aggregate

Expect of a group of HSPs. Hopefully, you’ve gained a better understanding of how

all of these operations of fit into the larger picture of Karlin-Altschul statistics.

You have also seen that it’s possible to use Karlin-Altschul statistics to recover statis-

tical measures that are calculated by BLAST internally, but not included in the

report—principally, sum scores and the individual Expect for an HSP for which an

Expect(n) has been reported. Learning to calculate these values is the first step

toward becoming a power user of BLAST statistics. The remaining sections of this

chapter will show you how to use what you’ve learned to deal with critical questions

about BLAST results.

Using Statistics to Understand BLAST Results

Karlin-Altschul statistics is much more than a way to determine the statistical signifi-

cance of a sequence alignment in the context of a database search. It also provides a

framework with which to probe the complex relationships that exist between BLAST

parameters and results. Using Karlin-Altschul statistics to ask and answer questions

about a BLAST search is much like using stoichiometry at the lab bench; it doesn’t

require theoretical savvy, just a little algebra. It’s also useful; you no longer need to

be frustrated when confronted with an inexplicable BLAST result.

Now let’s look at a practical application of Karlin-Altschul statistics: using BLASTN

to map a PCR primer to a genome. The application is a simple but striking example

of how to use Karlin-Altschul statistics to understand the way parameter choice

determines BLAST results. Finally, Karlin-Altschul statistics reveal much about

BLASTN’s strengths and weaknesses and its potential as a tool to detect the con-

served, cis-regulatory regions of genes.

Where Did My Oligo Go?

First, try to identify the position of the following oligo-nucleotide in the Drosophila

melanogaster genome using WU-BLASTN with its default parameters:

TACATCCGGCACTTAGCCGGGCTCG

Example 7-4 shows that the oligo isn’t found in the Drosophila melanogaster genome

that uses WU-BLASTN with default parameters.

This is the Title of the Book, eMatter Edition

Copyright © 2012 O’Reilly & Associates, Inc. All rights reserved.

110

|

Chapter 7: A BLAST Statistics Tutorial

There are, of course, many reasons why you might not be able to identify an oligo in

the Drosophila melanogaster genome. First, the oligo might contain repetitive

sequence and thus be masked out. However, because WU-BLAST doesn’t mask by

default, that can’t be the reason. Second, the assembled genome may be incomplete.

Every sequenced genome to date is incomplete to some degree. In fact, a 99 percent

complete 124mb genome is still missing 1.24 mega-bases of a euchromatic (nonre-

petitive DNA) sequence, leaving plenty of space for an oligo to go missing in. The

incompleteness of the genome is a possible explanation for our WU-BLAST result,

but is it the correct one? Before concluding that the oligo falls into a sequencing gap,

let’s try to run NCBI-BLASTN with its default parameters. Aha! The NCBI-BLASTN

results in Example 7-5 show that the oligo is present in the Drosophila melanogaster

genome and the HSP is assigned a significant Expect.

Example 7-4. The oligo isn’t found

Reference: Gish, W. (1996-2000) http://blast.wustl.edu

Notice: this program and its default parameter settings are optimized to find

nearly identical sequences rapidly. To identify weak similarities encoded in

nucleic acid, use BLASTX, TBLASTN or TBLASTX.

Query= oligo

(25 letters)

Database: na_whole-genome_genomic_dmel_RELEASE3.FASTA

7 sequences; 124,181,667 total letters.

Searching....10....20....30....40....50....60....70....80....90....100% done

Smallest

Sum

High Probability

Sequences producing High-scoring Segment Pairs: Score P(N) N

*** NONE ***

Example 7-5. Using NCBI-BLASTN to find the oligo

Sequences producing significant alignments: (bits) Value

2R 2R.3 assembled 23-11-2001 50 1e-06

X X release:2 length:21666217bp Assembled X chromosome arm seque... 32 0.25

3R 3R.3 32 0.25

U GenomicInterval:U 30 0.99

3L 3L.3 v.3e 23351213bp BCM HGSC guide:3l-mtp-eval.08apr02 28 3.9

2L 2L release:3 length:22217931bp Assembled 2L chromosome arm se... 28 3.9

>2R 2R.3 assembled 23-11-2001

Length = 20302755

Score = 50.1 bits (25), Expect = 1e-06

Identities = 25/25 (100%)