It is with some embarrasment that I announce yet another revison
of the FASTA 1.6 version - 1.6b2. Now available on uvaarpa.virginia.edu is:
fasta16b2.shar and fasta16b2.shar.Z
These versions fix a trivial problem with numbers in alignments (which
arose because fasta now allows negative offsets in alignments; this is
useful for aligning upstream 5'-flanking regions) and a somewhat more
serious problem which occured when one was optimizing every sequence
in a library.
The optimization problem appeared because for the first time,
doing a search with ktup=1 and optimizing every sequence in the
library takes only about twice as long as doing the search with ktup=1
and optimizing only a few sequences. (In the past, it usually took
about 5 times longer to do a fully optimized search). As a result,
for protein sequences, it is very practical to search an entire
database with ktup=1, optimizing every sequence (This takes about 15
min on a 25 Mhz Sun 4 with a 200 residue query sequence against PIR
release 30 - 10,000,000 residues in 34,000 entries). When I started
optimizing large libraries, I uncovered some bugs in how the best
scores were saved with optimization.
A FASTA search with ktup=1, optimizing all sequences, compares
very favorably with a rigorous Smith-Waterman search (ssearch, see
Pearson (1991) Genomics, 11:635) and takes about 1/7 as long with this
With luck, fasta16b2 will be stable for more than a few days.