kblin / glimmerhmm Goto Github PK

Copyright (c) 2003  by  Mihaela Pertea.

This system is licensed as open source under the terms in the file LICENSE.

GlimmerHMM system is currently trained for Arabidopsis thaliana, rice, and human. 
The system is already compiled for linux, 32 bit platforms. 
You can compile your own system for your unix platform just by typing 
make in the sources directory.

To run GlimmerHMM  just use your own compiled GlimmerHMM or the 
already existing pre-compiled GlimmerHMM from the bin directory.You will 
also need to specify the directory containing the training files.

For instance, if you want to run GlimmerHMM for Arabidopsis on a linux
platform you should type
>GlimmerHMM/bin/glimmerhmm_linux fasta.file -d GlimmerHMM/trained_dir/arabidopsis/

To train GlimmerHMM you should first run make in the train directory. 
Read readme.train from GlimmerHMM/train. After creating your own training
directory you can use it at the -d option with GlimmerHMM.

For human predictions, if you don't know the GC isochore of your input sequence, use the usual command:
>GlimmerHMM/bin/glimmerhmm_linux fasta.file -d GlimmerHMM/trained_dir/human
If you already know the isochore of where the sequence is comming use directly 
the training directory for that isochore. There were four isochores for which 
GlimmerHMM was trained for human: 0-43% GC, 43-51% GC, 51-57% GC, and 57-100% GC.
E.g. suppose the predicted isochore of the input sequence has a 39% GC 
content. The command to use for running GlimmerHMM should be:
>GlimmerHMM/bin/glimmerhmm_linux fasta.file -d GlimmerHMM/trained_dir/human/Train0-43

For additional information about GlimmerHMM see http://www.cbcb.umd.edu/software/glimmerhmm/.