BLAST

This is the Title of the Book, eMatter Edition

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

The BLAST Algorithm

|

81

all potential neighborhood words. The two-hit distance is set to 40 amino acids by

default, so words don’t have to be clustered as closely as they appear in the figures.

In principle, setting the two-hit distance to a smaller value also increases speed, but

in practice, its effect is insubstantial.

In WU-BLAST, you may set W to any value for any program. If W is set to 5 or more,

neighborhood word scores aren’t used; they are computed only by explicitly assign-

ing a value for T. High values of W in conjunction with moderate values of T can

lead to immense memory requirements, so it is best not to set T when W is 6 or

more. To alter the speed/sensitivity of WU-BLAST you can use a variety of combina-

tions of W, and T, and you can also employ the two-hit algorithm.

The statistical model underlying BLAST assumes the letters are independent of one

another so that the words MQV and MVQ have the same probability of occurring.

However, certain combinations occur in biological sequences much more often than

expected. These are usually low-complexity sequences—for example, FFF (see Chap-

ters 2 and 4). Low-complexity sequences are often of little biological interest and

aligning them wastes CPU cycles. Masking these regions is therefore common. Both

NCBI-BLAST and WU-BLAST ...