readme.v34t0

序列对齐 Compare a protein sequence to a protein sequence database or a DNA sequence to a DNA sequenc

 $Name: fa35_03_06 $ - $Id: readme.v34t0,v 1.172 2007/11/22 11:30:55 wrp Exp $

>>May 28, 2007

Small modification for GCG ASCII (libtype=5) header line.

>>January 12, 2007	fasta-34_26_2

Fix a problem with pssm_asn_subs.c reading strings (sequences) longer
than 1024 bytes.

Remove searchfa.cgi, searchnn.cgi, cgi-lib.pl, my-cgi.pl - this code
was used for an ancient FASTA WWW implementation and has been replaced
by the FASTA_WWW package.

FASTA Version numbers are being modified to make releases easier to
track, thus fa34t26b5 has become fasta-34_26_1.  I would prefer to use
decimal versions, but CVS does not allow '.' in tags.

>>January 4, 2007	fasta-34_26_1

Include scripts for building Mac OS X Universal binaries on a PPC
machine.  Programs are compiled first with Makefile.os_x (gcc-3.3 for
PPC) and then installed into ./ppc/.  Programs are next compiled with
Makefile.os_x86 for i386, and the resulting executables installed into
./i386/.  Finally, the "make_osx_univ.sh" script is run to build the
universal binaries from the two executables using "lipo".

>>December 12, 2006

Fix some problems with p2_workcomp.c: (1) no longer initialize pad
characters for non-existant sequences. (2) deal with small libraries
consistently with the serial versions.

>>November 17, 2006	fa34t26b5

Fixed a problem reading ASN.1 format 2 PSSM's.  It is now possible to
download a PSI-BLAST PSSM RID and search properly.  Next, the query
sequence from the PSSM should be used instead of the provided query
sequence, so that the query sequence is ignored.

>>October 19, 2006	fa34t26b4

Fixed problem with SSE2 code when PSSM's are used.

>>October 6, 2006	fa34t26b3

A new set of WIN32 programs is now available that use the Intel C++
9.1 compiler, rather than the much older Borland Turbo-C compiler. All
of the unthreaded programs that are part of the Unix and MacOSX FASTA
distributions are now available.  Threaded (multiprocessor) versions
of the program as available as well, as are sse2 accelerated versions
of ssearch34 (ssearch34sse2.exe, ssearch34sse2_t.exe).

Th new WIN32 code also uses Microsoft's "nmake" program to build the
programs, which allows much greater consistency between the Unix and
Windows versions.


>>September 18, 2006

Static global alignment variables removed from dropnfa.c, dropfx.c,
dropfz2.c.  dropnfa.c, dropfx.c and dropfz2.c should be thread safe.
Together with the earlier changes, all the FASTA functions should now
be thread safe during the alignment process.

>>August 17, 2006

Begin removal of static variables from Smith-Waterman alignment
functions.  These variables kept the functions from being thread-safe.
Now dropgsw.c and dropnsw.c are thread-safe.

>>August 15, 2006	fa34t26b2

Fixed a problem with pv34compfx/mp34compfx (and fy) producing
improperly labeled alignments and de-allocating memory for the reverse
complement.

>>July 18, 2006

The library file name parsing programs now provide the option for
environment variable substitions.  For example, SLIB2=/slib2 as an
environment variable (e.g. export SLIB2=/slib2 for ksh and bash), then

	fasta34 -q query.aa '${SLIB2}/swissprot.fa'  expands as expected.

While this is not important for command lines, where the Unix shell
would expand things anyway, it is very helpful for various
configuration files, such as files of file names, where:

	<${SLIB2}/blast
	swissprot.fa

now expands properly, and in FASTLIBS files the line:

	NCBI/Blast Swissprot$0S${SLIB2}/blast/swissprot.fa

expands properly.  Currently, Environment variable expansion only
takes place for library file names, and the <directory in a file of
file names.

>>July 14, 2006	  fa34t26b1

Updated Farrar smith_waterman_sse2.c code to address possible bug
(code from Michael Farrar).  Include <sunmedia_intrin.h> for
compilation with Sun compiler with Makefile.sun_x86.

>>July 2, 2006	  fa34t26b0

This release provides an extremely efficient SSE2 implementation of
the Smith-Waterman algorithm for the SSE2 vector instructions written
by Michael Farrar ([e-mail]).  The SSE code speeds up
Smith-Waterman 8 - 10-fold in my tests, making it comparable to Eric
Lindahl's Altivec code for the Apple/IBM G4/G5 architecture.

The Farrar code is largely confined to smith_waterman_sse2.c and
smith_waterman_sse2.h, which are copyright (2006) by Michael Farrar,
and cannot be redistributed without his permission.  Mr. Farrar has
agreed to provide his code under the same policy used by FASTA -
e.g. the code can be used without permission, but not redistributed.

The Farrar code uses GCC version 4.0 SSE2 intrinsic functions to avoid
assembly language code.  Unfortunately, in my hands, "gcc -O3" causes
"out of memory" errors, and other problems, so "gcc -O" is used instead.

>>June 23, 2006   fa34t25d10

Modifications to comp_lib.c, compacc.c, and other files to ensure that
function-specific MAXTOT values are used properly.  MAXTOT is now
available as m_msg.max_tot, which is set in initfa.c (m_msg.max_tot =
MAXTOT) to ensure that functions that need very large MAXTOT values
(e.g. TFASTX) can get them.  tfastx can now search successfully with
titin, a 27,000 residue protein.

Other changes have been made to accomodate long query sequences.

A serious bug was found in fastx34(_t) that caused alignment
coordinates to be calculated improperly when the DNA sequence was much
longer than the protein sequence.

>>May 31, 2006	fa34t25d9

Fixed some problems with fasts/fastf alignments when -m 9 options were
used.  Unlike the other algorithms, the a_res structure does not
capture all the information to re-produce an alignment, so do_walign
now sets bptr->have_ares to indicate whether the a_res structure is
valid.

Various problems with bad library names, and short query titles were
also fixed.

Updated version number/date on all drop*.c functions.

>>May 24, 2006	fa34t25d8

Revised code for NCBI *.pal/*.nal databases has been tested on all
architectures, including Windows.

In addition, support for ASN.1 PSSM:2 files provided by the NCBI
PSI-BLAST WWW site is included.  This code will not work with
iteration 0 PSSM's (which have no PSSM information).  For ASN.1
PSSM's, which provide the matrix name (and in some cases the gap
penalties), the scoring matrix and gap penalties are set appropriately
if they were not specified on the command line. ASN.1 PSSM's are type 2:
	ssearch34 -P "pssm.asn1 2" .....

>>May 18, 2006

Support for NCBI Blast formatdb databases has been expanded.  The
FASTA programs can now read some NCBI *.pal and *.nal files, which are
used to specify subsets of databases.  Specifically, the
swissprot.00.pal and pdbaa.00.pal files are supported.  FASTA supports
files that refer to *.msk files (i.e. swissprot.00.pal refers to
swissprot.00.msk); it does not currently support .pal files that
simply list other .pal or database files (e.g. FASTA does not support
nr.pal or swissprot.pal).

In the process of providing this support, the routines used to read
ASN.1 binary formatdb files were substantially improved.  It is now
possible to see multiple description lines for a single sequence.

IS_BIG_ENDIAN has been removed from all of the Makefiles.  The code
now looks for the definition of __BIG_ENDIAN__ or _BIG_ENDIAN to
decide whether the architecture IS_BIG_ENDIAN.  If, for some reason,
one of these macros is not defined on a BIG_ENDIAN architecture, then
-DIS_BIG_ENDIAN is required.

>>May 12, 2006	CVS fa34t25d7

Corrected serious problem with coordinate display calculation for
fasta34 and ssearch34 - in some cases the coordinates and alignment
symbols were off by the length of the context (typically 30 residues).

Added capability to read ASN.1 binary PSSM information.  This
information is provided (in an encoded form) from the NCBI PSI-BLAST
WWW site.  (What is actually provided from the WWW site is a bzip2-ed
binary file that is converted to ASCII HEX.  The ASCII HEX file must
be converted to binary, and then bunzip'ed. This bunzip-ed file is
binary ASN.1.)  These files can also be generated by 

 blastpgp -J T -C pssm.asn1_bin -u 2

I am parsing the ASN.1 binary manually, not using the NCBI toolkit, so
there may be some files that are not parsed properly - if so, let me
know.

(May 12, 2006 - The NCBI changed the format of the psi-blast ASN.1
PSSM - and has not yet provided documentation of the new structure, so
this code does not work. It does work with blastpgp v 2.2.13, but not
with the web site version 2.2.14.  A fix was provided 24-May-2006)

>>April 18, 2006

Small modification in mshowbest.c to provide more consistent display
widths with -m 9i in list of best hits.

>>April 11, 2006 CVS fa34t25d6

Corrected a problem introduced with the new, more efficient method for
displaying alignments.  For the tfast* programs, which must translate
the library sequence, translations were not done when alignments were
re-displayed.

Corrected an older problem with tfastx34 against very long sequence
databases - the code to more efficiently do the display alignment did
not use the correct sequence coordinates.

Modifications to dropfs2.c to ensure that exact peptide matches are
captured more frequently.

>>March 16, 2006 CVS fa34t25d5

Change to initfa.c to allow lower case DNA libraries using the
-DDNALIB_LC compile time option.

Modify p2_complib.c, p2_worklib.c (and doinit.c, msg.h) to allow the
-V annotation option for the parallel programs.  Also modify to allow
specification of the query range (but only for the first query, like
fasta34) for the parallel programs.

Modification of p2_workcomp.c to correct some problems presenting
percent similarity.  Also correct unreleased bugs in the alignment
routines that allow more efficient alignment re-calculation.

>>Nov 20, 2005

Changes to support asymmetric matrices - a scoring matrix read in from
a file can be asymmetric.  Default matrices are all symmetric.

>>Oct 24, 2005

Modifications extended to p2_complib.c/p2_workcomp.c.  Incorporation
of drop_func.h into p2_workcomp.c greatly simplifies things.  No
changes in communication - struct a_res_str is internal to
p2_workcomp.c.

Additional changes to do_walign() so that aln_func_vals() must be
called to set llfact, qlfact, etc in a_struct aln before or after
do_walign is called.  do_walign produces a_res_str a_res, which has
all the information necessary to produce a calcons() or calc_code()
alignment.

>>Oct 19, 2005 CVS fa34t26b0

Modifications to drop*.c and c_dispn.c to separate (and simplify) some
of the alignment coordinate calculations.  Before, the "a_struct" had
the coordinates of the alignment used in the display (seqc0, seqc1)
AND in the original sequences (aa0, aa1), as well as other information
used to calculate alignment coordinates.  In the new version, astruct
coordinates always refer to seqc0,1, while a new structure, a_res_str,
has coordinates for aa0, aa1 as well as the alignment encoding in res[nres].
Eventually, this should make it possible to display multiple local
alignments from the same two sequences.

In addition, the file "drop_func.h" has been added to the project, and
is included by many of the files (all the drop*.c functions,
mshowbest.c, mshowalign.c) to ensure that the various functions are
declared and used consistently.

>>Sept 19, 2005	CVS fa34t25d4

Changes to support Mac OS 10.4 - Tiger (include sys/types.h in more
files).  Documentation update for prss34/prfx34. Modifications to
comp_lib.c to support prss34_t/prfx34_t.  Shuffle numbers for
prss/prfx can now be specified by "-k #".

>>Sept 2, 2005

The prss34 program has been modified to use the same display routines
as the other search programs.  To be more consistent with the other
programs, the old "-w shuffle-window-size" is now "-v window-size".

prss34/prfx34 will also show the optimal alignment for which the
significance is calculated by using the "-A" option. 

Since the new program reports results exactly like other
fasta/ssearch/fastxy34 programs, parsing for statistical significance
is considerably different.  The old format program can be make using
"make prss34o".

>>Aug 26, 2005

Modifications to save_best() in comp_lib.c to support prss34_t.  It
did not work before.

>>July 25, 2005

Modify mshowbest.c to suppress gi|12345 in HTML mode.

>>July 18, 2005	CVS fa34t25d3

Modifications to Makefile.tc to support NCBI formatdb formats under
Windows.

>>May 19, 2005  CVS fa34t25d2

Modifications to dropfs2.c to fix an obscure bug that occurred when
correctly ordered peptides aligned one residue apart.

>>May 5, 2005 CVS fa34t25d1

Modification to the -x option, so that both an "X:X" match score and
an "X:not-X" mismatch score can be specified. (This score is also used

give a positive score to a "*:*" match - the end of a reading frame,
while giving a negative score to "*:not-*".

>>March 14, 2005  CVS fa34t25b4

Fixed some problems caused by padding characters required for
Smith-Waterman ALTIVEC in the parallel (p2_complib.c, p2_workcomp.c)
versions.

>>Feb 24, 2005	CVS fa34t25b3