Microarray assosiated motif analyzer Files

MAMA:Microarray associated motif analyzer
(Version 1.0)

MAMA is a name of method to predict cis-acting motifs.
MAMA require a set of Microarray data as an input.
MAMA uses SVM (Support vector machine) to simulate gene expressions based on predicted motifs.
MAMA selects the best parameter automatically, which maxmize the accuracy of simulation of gene expressions.

MAMA is availabbe at http://

Table of Contents
=================

-Requirements
-Installation
-Data preparation
-Usage

Requirements
============

 *CUDA compatible GPU cards
  It is strongly recommended to use Tesla or Quadro lines.
  If CUDA is not executable MAMA will be run without CUDA, but it takes several weeks to finish a calculation.
 *perl environment (and install Math-CDF-0.1)
  http://www.activestate.com/activeperl
  See http://www.edcallahan.com/math_cdf.html about Math-CDF-0.1
 *libSVM
  http://www.csie.ntu.edu.tw/~cjlin/libsvm/
 *gnuplot (This is required for libSVM python interface)
  http://www.gnuplot.info/
  Install gnuplot to "c:\tmp\gnuplot\binary\pgnuplot.exe" for windows, 
  or to "/usr/bin/gnuplot" for the others.
 *AUCCalculator 0.2
  http://mark.goadrich.com/programs/AUC/
 *python environment (This is required for libSVM python interface)
  http://www.python.org/
 *java environment (This is required for AUCCalculator 0.2)
  http://java.com
 *R environment (and install "igraph" package)
  http://www.r-project.org/
  Do not forget setting up PATH for perl, python, java and Rscript.
  

Installation
============

1. Move this MAMA folder to where you can access easily from command line (i.e. C:\MAMA).
   Read "Requirements" and prepare them.
   Check that you can execute perl, python, java and rscript from command line.
   
2. Copy easy.py and grid.py from libSVM/tools to MAMA/OptimizationTools folder.
   Copy libSVM/windows folder to MAMA for windows users.
   Copy svm-scale, svm-train and svm-predict executables to MAMA folder if not windows.
   Copy auc.jar (AUCCalculator 0.2) to OptimizationTools folder.
   
   
Data preparation
================
   
*If your microarray data was obtained from A. thaliana ATH1, just paste signal ratio to a template file ().
Inputfile is a csv format without " (Windows Microsoft Excel save this).
An inputfile should contain 3 columns "Gene id", "Sequence", and "Gene expression ratio".
"Gene id" should not contain characters like "/,\=&". The length of Gene id should be less than 4000.
"Sequence" should be 5000 bp sequence which contains 3000 upstream sequences from transcription start sites (TSS)
and 2000 bp downstream from TSS. If the sequence of a gene is shorter than 5000 bp or empty, this program will ignore the gene. 
The name of input file should contain "3k+2k".

Usage
=====
1. Move to this MAMA folder in command line ("cd C:\MAMA").
2. Execute "perl run_mama.pl <INPUTFILE>" from command line.
For example, execute "perl run_mama.pl promoter-ratio_3k+2k_ATH1_TAIR10_template.csv".

Results
=======
The predicted motifs will be listed in "motiflist/countmotifs4freq-list_output_500.csv" and "countmotifs4freq-list_output_150.csv".
"countmotifs4freq-list_output_500.csv" includes the name of predicted motifs from 500 bp upstream of the TSSs. 
"countmotifs4freq-list_output_500.csv" includes the name of predicted motifs from a region 50 bp upstream to 150 bp downstream of the TSSs. 
The logo of predicted motifs will be generated in "motifgraph" folder.
Candidate motif pairs that co-regulate genes will be listed in "OptimizationTools/comma_output_500+150_ranking_clean.csv".
Currently, this software does not automatically evaluate "highly possible" and "possible" motif pairs. 
"OptimizationTools/best_mama_parameters.txt" contains the best Nmp (Top Nmp of motif pairs are highly possible candidate).
Open "OptimizationTools/progress_optimizemama_parameters.csv" and extract "NmpX,AUC-ROC = Y.YYYYYYYYYYYYYYYYYYYY". X is a value of Nmp. Therefore calculate P value by comparing between AUC-ROCs of Nmp0 and the others using student t-test. If P value < 0.05, Top Nmp of motif pares are possible candidate.