Written by: Maria Chatzou & Pantelis Bagos
Briefly, MIGMA is based on a Mixture Transitiom Distribution model where each lag contributes to the prediction of the current letter in a separate and additive way.
For a more detailed description, as well as literature pointers please see the application note under the doc/ directory, which accompanies the code.
Compile MIGMA using "make" Delete MIGMA executable using "make clean" Delete MIGMA executable and object files using "make wipe"
Both training and test sequence files must be in FASTA format (see http://tigrblast.tigr.org/web-hmm/fasta.html ).
In the example/ directory you will find the following files:
inputs: 7tm_3 - a family from Pfam, which we will be used for prediction 7tm_3_TRAIN - 4/5 of the family sequences, which will be used for training
(you can copy these to a new directory, say example2, and compare results)
Training
Gerneral:
./migma -itrain <input_training_file> -om <mtd_output_file> -a <alphabet (1:protein, 2:DNA)> -alg <algorithm (default: EM, 1: EM, 2:Viterbi)> -or <order>
(in the example/ directory, type the following command line)
In Linux:
../migma -itrain 7tm_3_TRAIN.fa -om 7tm_3.mtd -a 1 -or 20 -alg 1
In Windows:
../migma.exe -itrain 7tm_3_TRAIN.fa -om 7tm_3.mtd -a 1 -or 20 -alg 1
yields the following output files:
7tm_3.mtd - the MTD file that results from training from 7tm_3_TRAIN.txt
Prediction
General:
./migma -m <mtd_file> -i <input_file> -o <output_file>
(in the example/ directory, type the following command line)
In Linux:
../migma -m 7tm_3.mtd -i 7tm_3.fa -o 7tm_3.pred
In Windows:
../migma.exe -m 7tm_3.mtd -i 7tm_3.fa -o 7tm_3.pred
yields the following output file:
7tm_3.pred - the results of the prediction
option:
-p when used it also yields a second output file which contains the predisctions of the amino acid probabilities
Training & Prediction:
General:
./migma -i_train <input_training_file> -o_mtd mtd_output_file> -a <alphabet (1:protein, 2:DNA)> -alg <algorithm (default: EM, 1: EM, 2:Viterbi)> -or <mtd_order> -i <input_file> -o <output_file>
(in the example/ directory, type the following command line)
In Linux:
../migma -i_train 7tm_3_TRAIN.fa -o_mtd 7tm_3.mtd -a 1 -or 20 -alg 1 -i 7tm_3.fa -o 7tm_3.pred
In Windows:
../migma.exe -i_train 7tm_3_TRAIN.fa -o_mtd 7tm_3.mtd -a 1 -or 20 -alg 1 -i 7tm_3.fa -o 7tm_3.pred
yields the following output files:
7tm_3.mtd - the MTD file that results from training from 7tm_3_TRAIN.txt
7tm_3.pred - the results of the prediction using 7tm_3.mtd as the model
Maria Chatzou e-mail: maria.chatzou@crg.eu