VERSION 1.20

* Moderately improved start site prediction by adding an upstream
scoring system for the -1/-2 positions and the -15 to -45 positions.
* Added a routine to correct starts of rarer types (TTG in E. coli, etc.)
that win via every other parameter (coding, RBS, and upstream score).
* Reduced the minimum gene size for genes running off edges to 50bp.  This
should help in identifying fragmented genes in draft or metagenomic data.
* Added more detailed information to the starts file, including ATG/GTG/TTG
usage, the RBS bin and spacer distance, and the new upstream composition
score.
* Added a '-n' option to force the program not to use the SD motif finder.  
(This can be useful if you suspect the organism uses a motif like AAGAGG and 
you would like to get to the AAG/etc. motifs which will be missed by the SD 
finder.  It is also useful to get very specific RBS motifs and spacers in
the starts file).
* Added gene.c/h files containing a data structure for final gene models
(exported from the dynamic programming).  Moved printing of translations and
gene models to this data structure.
* Added GFF output specifiable with the '-o gff' option.
* Fixed a bug where running with a training file and without a training file
produced different results (this elusive bug has existed since 1.0).
* Modifed intergenic score weighting and fixed some minor bugs with this
function.
* Fixed some bugs that only occurred when compiled for Windows.
* Modified draft_prodigal.pl to catch signals and generally run more robustly.

Starts files are no longer in the same format as 1.0-1.10.
Training files for 1.0-1.10 are NOT compatible with 1.20+.

VERSION 1.10

* Added support for multiple FASTA input via a wrapper script called
"draft_prodigal".  Requires Perl 5.
* Added an alternative start scoring model that automatically discovers RBS
motifs and builds a model from upstream regions.  This model was found to
perform slightly worse in SD genomes than the existing code, and it also takes
twice as long, so we only use it when a genome is determined not to use the SD
motif.
* Fixed a bug in parsing Genbank files where the parser did not understand lines
with "[gap 100] Extend Ns" etc. in them.

Training files for 1.0-1.05 are NOT compatible with 1.10+.

VERSION 1.05

* Added support for exclusion of RNA regions from gene modeling via masking such
objects in the sequence with n's and using the '-m' option to make Prodigal
recognize them as such.

VERSION 1.04

* Modified Prodigal to allow genes to run off edges.  (User can disallow this
with the -c option).
* Allowed Prodigal to run on any size sequences if using data from a training
file.

VERSION 1.03

* Fixed a bug where Prodigal didn't handle Windows C/R's in sequence input.
* Modified the start trainer to function better in the absence (or scarcity) of
good genes to train on.

VERSION 1.02

* Added protein translation support via -a flag.
* Added full translation table support via -g flag and removed the -n option.
* Fixed a bug that caused prodigal to hang occasionally (involving 4 consecutive
overlapping genes in the dynamic programming model).

Training files for version 1.01 are NOT compatible with version 1.02+.

VERSION 1.01

* Reduced the number of false positives by fixing a bug in the
eliminate_bad_genes routine that was causing genes not to be omitted.
* Increased the accuracy of the RBS scorer slightly by creating more bins based
on distance and motifs and ordering then slightly differently.
* Fixed a bug in the dynamic programming model in zero gene cases.

Training files for version 1.0 are NOT compatible with version 1.01.
