BLAST

This is the Title of the Book, eMatter Edition

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

BLASTN Protocols

|

141

Clustering with blastclust

Given a database of DNA sequences, it is often necessary to rapidly group related

sequences for further analysis or simply identify redundancy at some level. One

approach is to use BLASTN with rapid, insensitive search parameters, and then parse

the output for the desired properties (e.g., 97 percent identity over at least 90 per-

cent of the sequence length), and finally group all reads that are directly or indirectly

(transitively) associated. Bioperl tools can automate such a procedure, but it takes a

little work. The NCBI-BLAST distribution includes a standalone program called

blastclust that is designed for just this task.

Approach

Two protocols are given below—one for clustering ESTs that are expected to be

nearly identical across the length of the read (99 percent identity, 90 percent cover-

age), and another for shotgun sequences that have high identity over a smaller region

of the read (97 percent identity, 10 percent coverage). The alignment parameters are

preset for near identity, but some differences that may be the result of sequencing

errors or polymorphism are allowed. Unlike other BLAST programs, blastclust

doesn’t allow soft masking.

EST clustering

blastclust –i <fasta file> -o <output file> -p F -L 0.9 –S 99 –b F

Shotgun sequences ...