***************************************************
 PconsC : Combination of direct information methods 
         and alignments improves contact prediction
***************************************************

PconsC requires:
	* Python interpreter (tested from Python 2.6 onwards)
	* jackhmmer (part of HMMER package http://hmmer.janelia.org/software)
	* HHblits (ftp://toolkit.genzentrum.lmu.de/pub/HH-suite/)
	* PSICOV (http://bioinfadmin.cs.ucl.ac.uk/downloads/PSICOV/)
	* sequence databases for jackhmmer (BLAST-formatted protein database, 
	  e.g. nr90, uniref90) and HHblits (bundled with software, nr20)
	* MATLAB for running plmDCA (necessarily with Java VM enabled)

After installing the necessary software, set paths in localpaths.py

Usage:
------

PconsC to work requires 16 input files: 
	8x PSICOV predictions for jackhmmer and HHblits alignments at 4 e-values
	8x plmDCA predictions for jackhmmer and HHblits alignments at 4 e-values

The method is optimized for e-value cutoffs of: 1e-40, 1e-10, 1e-4 and 1e0.

If you have these files, use ./predict.py to predict contacts

PconsC is capable of generating these file on its own, provided the proper paths
are set. Use ./predictAll.py to run all the predictions and 

Output format:
aa1 	aa2 	contact_propensity

Example:
1	5	0.003
1	6	0.007
1	7	0.200
[...]
27	43	0.952

The higher the propensity, the more likely are two residues to be in contact
