2.1.1 
-----------------------------------------------------------
Initial Public Release

2.1.2 
-----------------------------------------------------------
Includes additional parsing modules to read different formats.
Check the README file to see what formats are supported.

2.1.2a  
-----------------------------------------------------------
Bug fixed in training procedure.  Coding vectors on the opposite
strand of the true gene were being mislabeled as coding.

2.1.3 
-----------------------------------------------------------
Allows the use of partial gene models in the input for training

2.1.4 
-----------------------------------------------------------
Intron Penalty length, fixed Gff reading error

2.1.5  
-----------------------------------------------------------
Compiles on both GCC 3.2.2 (still with deprecated flags) and 2.95 
more OS and compile platforms to come
Read the "snap" gene prediction file format

3.1.1 
-----------------------------------------------------------
Base line for a new release version, which allows for new 
variable length sequence interval parameters.  This version
includes no new parameters, and should have identical output
to version 2.1.5, however, due to algorithmic changes the
program will run slower.  The drop in runtime speed will
hopefully be made up for with more accurate predictions
in future versions.

Note this version is not expected to be released, since
there is no difference in output and 
additional code changes needed to streamline
the code.

3.1.1a 
-----------------------------------------------------------
Cosmetic code changes mainly, however, a few minor bug
fixes detected over 3.1.1 - This version is being
tested on the 515 Arabidopsis BACs to ensure identical
output with 2.1.5, so far it appears that there
were a few bugs in 2.1.5 that the new algorithm fixes

3.1.3
-----------------------------------------------------------
Can use an intron prediction model (as an option)
and shows an improvement over the published results.

3.2.0
-----------------------------------------------------------
Intron prediction model fully tested. Now we assume that 
Intron data is explicitly defined in sequence alignment data
by requiring that multiple exons created from a single source
cDNA are identified in the input file. This version is tested
on Human and Rice.
3.2.2
-----------------------------------------------------------
Bug fix, sequence alignment data was not properly sorted
by coordinates - it assumed the data was already sorted.

3.2.6
--------------------------------------------------------
Linear combiner option added.
