readme.v34t0

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

but nothing that is dependent on aa0.  The aa0 dependent stuff (nm0,
Lambda, K, etc) is now stored in struct mngmsg.  This was mostly done
to support the pv34comp* programs, which have separate mngmsg
structures but the same pstructs.

The fasts34, fasts34_t, and pv34compfs/c34.workfs have all been tested
successfully.

>>July 19, 2002

Fix an old bug in the calculation of E()-values in DNA databases
longer than 2147483647 residues on machines with 32-bit longs.


>>July 28-31, 2002

(1) The various Makefile's have been "normalized".  The fast*34[_t]
    (Makefile.34m.common[_sql]), Makefile.pvm4[_sql], and
    Makefile.mpi4[_sql] make files all use a common set of filenames,
    described in Makefile.fcom.  This greatly simplifies adding
    programs, but requires that all *.o files be deleted when moving
    from fast*34* to pv34comp* to mp34comp*.

(2) showalign.c/p_showalign.c have been merged into mshowalign.c
    showbest.c/manshowbest.c have been merged into mshowbest.c.  Some
    of the related files (showun.c, manshowun.c, have not been merged
    or tested).

(3) Code for ranking scores with valid e_value's incorporated.

(4) Bug fixes in p2_complib.c, so that fasts34/fasts34_t/pvcompfs
    provide identical statistics.

>>July 26, 2002

Makefile.pvm4_sql and Makefile.pvm4 have been substantially simplified
by providing the worker program name from the h_init() function in the
initfa.c/initsw.c files.

>>July 24, 2002

Substantial modifications to param.h, structs.h to ensure that no
sequence specific information is kept in struct pstruct.  This
structure now holds the pam[] matrix, and other scoring parameters,
but nothing that is dependent on aa0.  The aa0 dependent stuff (nm0,
Lambda, K, etc) is now stored in struct mngmsg.  This was mostly done
to support the pv34comp* programs, which have separate mngmsg
structures but the same pstructs.

The fasts34, fasts34_t, and pv34compfs/c34.workfs have all been tested
successfully.

>>July 8, 2002

Modifications to comp_lib.c, initfa.c and new scaleswt.c, tatstats.c
to support FASTS with Tatusov statistics.

last_params() has been introduced to allow aa0 dependent changes in m_msg/pstr.

sortbest() has been moved into initfa.c/initsw.c to make it function specific.

find_z() takes an additional parameter, escore.

The do_work() results structure, beststr, and stat_str all accommodate
escores as well as integer scores (stat_str also saves segn and segl
but doesn't need them).

In scaleswt.c, process_hist() now knows much more about Tatusov statistics.

last_stats() provided to accommodate rank-based statistical corrections.

scale_scores() is the last function to modify the beststr scores
(final calculation of E-value).

Some sortbest*() calls and some bptr[i]->zscore=find_zp() loops have
been moved into scale_scores();

>>July 3,5, 2002

Modifications to allow mySQL comments (--) in "library.sql 16" files.
Thus, a first line of:

	--host seqdb user password;

is read by FASTA as the login information to a mySQL server, but is
ignored by mySQL.  "DO" commands in FASTA mySQL files can also be
rendered invisible to mySQL in this way.  See "do.sql".

Modifications to mysql_lib.c to allow very long SQL statements.  The
buffer is now dynamically reallocated in 4Kb chunks.

The fasta3.1 man page has been updated and re-organized.

>>June 26, 2002

Minor modifications to nmgetaa.c (openlib()) to use the same arguments
for searching and PRSS.  PRSS needs access to all of m_msg, but
searches do not.  Other small fixes to comp_mlib.c, towards the goal
of merging comp_mlib.c and comp_lib.c.

>>June 25, 2002

Modify the statistical estimation strategy to sample all the sequences
in the database, not just the first 60,000.  The histogram is still
based only on the first 60,000 scores and lengths, though all scores
an lengths are shown.  The fit to the data may be better than the
histogram indicates, but it should not be worse.

Currently, this modification is available only if the -DSAMPLE_STATS
option is defined.

>>June 23, 2002	CVS fa34t11d4

Fix a very long-standing bug in fasty/tfasty that caused 'NNN' to be
translated as 'S', rather than 'X'.  fastx/tfastx has done this
correctly for many years, but the fasty/tfasty code that I received
from Zheng Zhang was not implemented correctly (my fault, his code was
fine).

>>June 19, 2002

Added "-C #" option, where 6 <= # <= MAX_UID (20), to specify the
length of the sequence name display on the alignment labels.  Until
now, only 6 characters were ever displayed.  Now, up to MAX_UID
characters are available.

>>May 30, 2002	CVS fa34t11d3

Fixed problem with programs using the default -E cutoff when -b was
provided.  With this implementation, -E can override -b, but -b
overrides the default -E.

Fixed problem with 64-bit file offsets in param.h (change USE_FSEEK0
-> USE_FSEEKO, include -D_LARGEFILE_SOURCE and -D_LARGEFILE64_SOURCE
in Makefile.linux_sql).  Put limits on alignment display length (200
chars).  More checks for null returns from SQL queries.

>>Apr 17, 2002	CVS fa34t11d2

Fixed bug in mm_file.h/ncbl2_mlib.c that caused the SGI version to be
unable to read blast2 format files.

Changed "mp_*" tags to "pg_*" for -m 10 option.

>>Mar 30, 2002

Fix embarrassing bug in revcomp() (getseq.c) that failed to complement
the central nucleotide in a sequence with an odd number of residues.

Small changes to dropfs.c for more segments.

>>Mar 16, 2002

Added create_seq_demo.sql, nt_to_sql.pl to show how to build an SQL
protein sequence database that can be used with with the mySQL
versions of the fasta34 programs.  Once the mySQL seq_demo database
has been installed, it can be searched using the command:

	fasta34 -q mgstm1.aa "seq_demo.sql 16"

mysql_lib.c has been modified to remove the restriction that mySQL
protein sequence unique identifiers be integers.  This allows the
program to be used with the PIRPSD database.  The RANLIB() function
call has been changed to include "libstr", to support SQL text keys.
Due to the size of libstr[], unique ID's must be < MAX_UID (20)
characters.

A "pirpsd.sql" file is available for searching the mySQL distribution
of the PIRPSD database.  PIRPSD is available from
ftp://nbrfa.georgetown.edu/pir_databases/psd/mysql.

>>Mar 6, 2002

Fix showbest.c showbest() to report pst.zdb_size as database size.
Fix dropnfa.c spam() to address off-by-one on end of run, and double
counting on backwards scan.  Fix dropnfa.c do_fasta() to fix another
problem introduced by -S.  Changes to comp_lib.c to ensure that both
the beginning and end of the query and library sequence have '\0'
present.  Changes to initfa.c, initsw.c to ensure that a match to a
lower-case letter with -S gets exactly the same score as a match to an
'X'.  Changes to mmgetlib.c to work with 64-bit longs in *.xin files.

>>Feb 26, 2002

Fixes to doinit.c, initfa.c, initsw.c to allow DNA matrices using the
"-s dna.mat" option.  A new matrix, "d50ry.mat" is available that
scores +5 for a match, -2 for a transition, and -5 for a
transversion. "d50ry.mat" corresponds to DNA PAM50 with transitions
twice as common as transversions.  When "-s dna.mat" is used, "-n"
MUST be used as well.

Query sequence names ("aa", "nt") should be more accurate.

>>Feb 22, 2002

Fix to getseq.c to allow "plain" sequence files.

>>Feb 12, 2002

Minor fix to res_stats.c.

>>Jan 28, 2002

Fixes to resurrect res_stats.c.  res_stats (cc -o res_stats
res_stats.c scaleswn.c -lm) takes the output from a current "-R
file.res" file and calculates statistical significance - this allows
one to take exactly the same set of scores (and lengths) and calculate
statistical estimates using different strategies.

>>Jan 24, 2002

modifications to mmgetlib.c, ncbl2_mlib.c to more robustly read memory
mapped files (*.xin, map_db) on machines lacking "native" 64-bit
longs.  If the machine provides some definition for a 64-bit long
(e.g. "long long", "int64_t"), things should work. 64-bit offsets into
memory mapped files work properly on Alpha, SGI, i386 Linux, and
MacOSX.  The current implementation depends either on 64 bit longs
(Compaq Alpha's pre 4.0G) or the <sys/inttype.h> file.  Makefile,
Makefile.alpha, and Makefile.linux have been modified.

Modifications to nmgetlib.c, mmgetlib.c to provide GI numbers and
Accession versions for Genbank searches.  If the GI:123456 number is
available, it will be used and the description line will be formatted:

	gi|123456|gb|ACC1234.1|LOCUS description

This should help FAST_PAN runs, where the version of a sequence
changes frequently.

>>Jan 10, 2002

Modifications to p2_complib.c, p2_workcomp.c to more reliably allocate
space for library sequence descriptions on the master and workers.

>>Jan 2-3, 2002		CVS fa34t10c/fa34t10d3

Fixes to comp_lib.c to support Macintosh and Windows/Turbo-C
compilation.  New Makefile.tc.  Macintosh version supports both
"Classic" and "Carbon" environments.

"<values.h>" has been replaced with the more modern "<limits.h>"

Fixes to p2_complib.c to support n_libstr (libstr length) in GETLIB().

comp_thr.c, complib.c removed.

>>Dec 16, 2001

Complete integration of comp_mlib.c with both the unthreaded and
threaded programs.  Comp_mlib allows fasta34 and fasta34_t to compare
a database with a second database, just as pv34compfa does.  Using
multiple queries with fasta34_t is not as efficient as pv34compfa (and
it cannot use networks of Unix workstations), but it is much easier to
use and install.

With the comp_mlib.c option, fasta34 cannot automatically recognize
DNA sequences, just as pv34compfa no longer recognizes DNA sequences.
You must use the "-n" option to search with DNA sequences.  The other
programs (fastx34, tfastx34, etc) "know" the type of the query and
database sequences, so "-n" is only required for fasta34(_t).

>>Dec 14, 2001		CVS tag fa34t10b

Fix problems reading DNA databases in blast2 format.

>>Dec 11, 2001

Changes to spam() in dropnfa.c so that, for DNA sequences, the
previous behavior for finding the boundaries of a local alignment
region use the same algorithm as previous versions of fasta.  For
protein sequences, the algorithm will extend the local region beyond
the "ktup" boundaries if a better score can be found.  For DNA
sequences, this raises the noise rather than increasing sensitivity,
so it is turned off and "ktup" boundaries are respected.  The old,
"ktup" boundary algorithm is available with -DNOSPAM_EXT.

This version also includes a working res_stats.c, which can be used to
test various statistical estimates on exactly the same set of scores.

Fixed problems with -m 9 percent identity for fastx/fasty/tfastx/tfasty.
These errors have been present since -m 9 was implemented.

>>Dec 10, 2001

Fix to map_db.c to work correctly with files > 2 Gb when 64-bit longs
are available.  It is not yet designed to work with ftello() and other
offset types.

>>Nov 11,21, 2001	CVS tag fa34t10a, fa34t10d1

Substantial changes to revcomp(), getseq(), and other functions to
correct problems with -S on DNA sequences.  Sequences with lower case
nucleotides were not recognized or reverse complemented properly.

Fix to dropnfa.c (v34t07, Nov 21, 2001) bg_align() to re-initialize
static globals - this fixes a problem encountered with pv34compfa.  A
new main program, comp_mlib.c has been added to the CVS archive,
although it is not referenced in any of the Makefile.  comp_mlib.c
works like p2_complib.c and compares a library against another
library.

>>Nov 4, 2001

Change to dropnfa.c spam () while(1) -> while(lpos <= dmax->stop).
This fixes a problem with ktup=1 on Suns only, so far.

>>Oct 4, 2001		CVS tag fa34t10

Add comp_lib.c file, which merges complib.c (unthreaded) and
comp_thr.c (threaded) code into one file.

Modifications to nmgetlib.c, mmgetaa.c to allow Genbank flatfile
format without DESCRIPTION or ACCESSION lines.

Additional fix for -S with ktup=1.

>>Sept. 24, 2001

Fix to have correct gap-penalties for short scoring matrices with
tfastx/fastx.

>>Sept. 10, 2001	CVS tag fa34t05d6

Fix a bug introduced by -S fix in fa34t05d5.  Also, try to remove
changes in p34compfa compared to pv4compfa output.

>>Sept. 6, 2001		CVS tag fa34t05d5

Fix the -S dropnfa/fx/fz2 bug that was not actually fixed in
fa34t05d4.  Incorporate the correct scaleswn.c refered to in
fa34t05d4.

>>Sept. 5, 2001		CVS tag fa34t05d4

Fix problem with m_msg.quiet that prevented interactive prompts for
ktup, file name, etc with threaded programs.