Initial commit

read_me.md

+### FilterDCA 
+### interpretable supervised contact prediction using inter-domain coevolution
+
+FilterDCA used 2 features to compute a probability of being a contact for a cople (i,j) in domain1 and domain2.
+The first feauture is the result of the method plmDCA
+The second one is a pattern score wich can be computed using the script and the maps given.
+
+To use the script you need:
+- the result of plmDCA for the join-MSA of the 2 domains ;
+- the lengh of the 2 domains ;
+- the size of filters you want to use : 5, 13, 21, 37, 45 or 69 ;
+- and to set the size of the M effictive ('medium' if under 200 and 'big' otherwise).
+
+In the 2 folders you can find:
+- the 6 maps (3 corresponding to helix-helix contact, and 3 for strand-strand contacts) for each of the possibles size (5, 13, 21, 37, 45 or 69)
+- the classifier, and the 'min' 'max' and value to normalise the correlation/ pattern score
+
+You can then produice :
+- the pattern score: the best correlation score matrix
+- the matrix of probabililty of contact
+- and finaly the predicted contact map