>>July 22, 1998

 --> v31t10

Corrected problem with histogram when unscaled statistics used (e.g. prss3).

Corrected problems with prss3 shuffled sequence prompt.  Provided option
to enter number of shuffles, window size, for prss3.  Number of shuffles
for prss3 can be entered as an option (-d #) or as the third argument
on the command line (prss3 query lib 1000).

Modified nrand.c, nrand48.c to use time to set random number.

Corrected problems reading GCG formatted files with prss3.

Corrected various problems with pvcomp* programs, but they still do
not produce alignments with version 3.1.

Two new programs, fastf3(_t) and tfastf3(_t) are available.  These
programs compare a set of mixed peptide sequences from an Edman
sequencer to a protein (fastf3) or DNA (tfastf3) database, using 
the database sequences to de-convolve the peptide mixture.

See fastf3.1

>>August 11, 1998

(no version change)

Modified initfa.c so that using '-n' on the fastx/fasty command line
would not cause problems.

Changed labeling of query sequence length for fastx/fasty from 'aa' to 'nt'.

>>August 18, 1998

(no version change)

Modified complib.c, comp_thr.c scaleswn.c, to report E()-value for only
one related sequence if -z 3 is used.

>>August 23, 1998

 -->v31t11

Some serious problems with prss3 have been corrected:

(1) use dropnsw.c rather than dropgsw.c for more accurate low scores

(2) modify estimation program; use scaleswe.c rather than scaleswn.c.
    scaleswe.c has some improvements for estimation by moments and can
    use MLE as well as mu/var (-z 3).

(3) add p() estimate.

(4) correct bugs in nrand48, which caused bad sequences for llgetaa.c

(5) -Z number works properly for prss3 and other programs (fixed histogram).

(6) a new program, ssearch3e, is available that uses the same scaling
    routines as prss3 (scaleswe.c). prss3 will save the random
    sequences it generates when the -r file option is given; the
    sequences are in file_rlib.  ssearch3e (or ssearch3 or fasta) can
    then do a search on exactly the same sequences that were used by prss3.

A bug reading GCG format compressed DNA databases was fixed.

Fixed a bug that caused query sequence not to be displayed with -m 10.

Simple optimization in dropnfa.c improves performance 10%.

>>Sept. 1, 1998

(no version change)

Modified nxgetaa.c to recognize "ACGTX" as nucleotides.

>>Sept. 7, 1998

 --> v31t12

Added -z 11 - 15, which use shuffled sequences, rather than real
sequences to calculate statistical estimates.  Because a shuffled
sequence score is calculated for each sequence score, the search
process takes twice as long.  In this first version, codons are not
preserved during shuffles, so tfasta/x/y shuffles may not be as
informative as they should be.

Also fix a problem with prss3 shuffles.

>>Sept. 14, 1998

 (no version change; previous version not released)

Corrected bugs in tfastx3/tfasty3 caused by using the -3 option with
or without -i.  With the bug fixes; "-3" and "-3 -i" work as expected;
"-3" gives the forward three frames, while "-3 -i" gives the reverse
three frames.

In addition, tfasta3/tfasta3_t was upgraded to perform the same way
that tfastx/y3 does - i.e. a search with "-i -3" searches only frames
4,5, and 6, while "-3" searches only frames 1, 2, and 3.

>>Sept. 29, 1998

 --> v31t13

Corrected bugs in dropfx.c that were corrected in fasta30 last May,
but lingered in fasta31.  Also included code to ensure that tfastx/y
alignments against long introns would not overrun the alignment
buffer.  Instead of overrunning the buffer, the message: ***aligment
truncated *** is displayed.