diff --git a/multi-ego-basic.ff/ffbonded.itp b/multi-ego-basic.ff/ffbonded.itp
index 4fe939f0..c64d685e 100644
--- a/multi-ego-basic.ff/ffbonded.itp
+++ b/multi-ego-basic.ff/ffbonded.itp
@@ -536,7 +536,7 @@
 #define gd_44K  -120.000 1.00 1
 ; Backbone dihedral angle -C-N-CA-C- LYS
 ;
-#define gd_45K 180.000 2.00 1
+#define gd_45K 180.000 1.40 1
 ; Backbone dihedral angle -N-CA-C-N- LYS
 ;
 ;;;;;;;;;;;;;;;;;;;;
diff --git a/src/multiego/ensemble.py b/src/multiego/ensemble.py
index 29e7d6c0..61a33a8b 100644
--- a/src/multiego/ensemble.py
+++ b/src/multiego/ensemble.py
@@ -349,6 +349,10 @@ def init_meGO_matrices(ensemble, args, custom_dict):
         reference_contact_matrices[name] = io.read_molecular_contacts(
             path, ensemble["molecules_idx_sbtype_dictionary"], reference["reference"], path.endswith(".h5")
         )
+
+        # TODO: probabilities of equivalent atoms should be averaged
+
+
         reference_contact_matrices[name] = reference_contact_matrices[name].add_prefix("rc_")
         reference_contact_matrices[name]["c6_i"] = [lj_data_dict[x][0] for x in reference_contact_matrices[name]["rc_ai"]]
         reference_contact_matrices[name]["c6_j"] = [lj_data_dict[x][0] for x in reference_contact_matrices[name]["rc_aj"]]
@@ -1035,7 +1039,7 @@ def set_sig_epsilon(meGO_LJ, parameters):
     # by comparing them with RC_probabilities so that the resulting epsilon is between eps_min and eps_0
     condition = (
         meGO_LJ["probability"] > meGO_LJ["limit_rc_att"] * np.maximum(meGO_LJ["rc_probability"], meGO_LJ["rc_threshold"])
-    ) & (meGO_LJ["probability"] > meGO_LJ["md_threshold"])
+    ) & (meGO_LJ["probability"] > meGO_LJ["md_threshold"]) & (meGO_LJ["epsilon_prior"] > 0.)
 
     meGO_LJ.loc[condition, "epsilon"] = np.maximum(0.0, meGO_LJ["epsilon_prior"]) - (
         (meGO_LJ["epsilon_0"] - np.maximum(0.0, meGO_LJ["epsilon_prior"])) / np.log(meGO_LJ["rc_threshold"])
@@ -1043,6 +1047,22 @@ def set_sig_epsilon(meGO_LJ, parameters):
     meGO_LJ.loc[condition, "learned"] = 1
     meGO_LJ.loc[condition, "sigma"] = meGO_LJ["distance"] / 2.0 ** (1.0 / 6.0)
 
+    condition = (
+        meGO_LJ["probability"] > meGO_LJ["limit_rc_att"] * np.maximum(meGO_LJ["rc_probability"], meGO_LJ["rc_threshold"])
+    ) & (meGO_LJ["probability"] > meGO_LJ["md_threshold"]) & (meGO_LJ["epsilon_prior"] < 0.) & (meGO_LJ["rc_probability"] > meGO_LJ["md_threshold"])
+    meGO_LJ.loc[condition, "epsilon"] = (-meGO_LJ["rep"] * (meGO_LJ["distance"] / meGO_LJ["rc_distance"]) ** 12).clip(
+        lower=-20 * meGO_LJ["rep"], upper=-0.05 * meGO_LJ["rep"]
+    )[condition]
+    meGO_LJ.loc[condition, "learned"] = 1
+
+    #condition = (
+    #    meGO_LJ["probability"] > meGO_LJ["limit_rc_att"] * np.maximum(meGO_LJ["rc_probability"], meGO_LJ["rc_threshold"])
+    #) & (meGO_LJ["probability"] > meGO_LJ["md_threshold"]) & (meGO_LJ["epsilon_prior"] < 0.) & (meGO_LJ["rc_probability"] <= meGO_LJ["md_threshold"])
+    #meGO_LJ.loc[condition, "epsilon"] = -meGO_LJ["rep"] * (
+    #   1.0 + (np.maximum(meGO_LJ["rc_probability"], meGO_LJ["rc_threshold"]) - meGO_LJ["probability"])
+    #)
+    #meGO_LJ.loc[condition, "learned"] = 1
+
     # Not-attractive interactions
     # this is used only when MD_th < MD_p < limit_rc_att*RC_p
     # negative epsilon are used to identify non-attractive interactions
@@ -1051,14 +1071,21 @@ def set_sig_epsilon(meGO_LJ, parameters):
         & (meGO_LJ["probability"] > meGO_LJ["md_threshold"])
         & (meGO_LJ["rc_probability"] > meGO_LJ["md_threshold"])
     )
-    # meGO_LJ.loc[condition, "epsilon"] = -meGO_LJ["rep"] * (
-    #    1.0 + (np.maximum(meGO_LJ["rc_probability"], meGO_LJ["rc_threshold"]) - meGO_LJ["probability"])
-    # )
     meGO_LJ.loc[condition, "epsilon"] = (-meGO_LJ["rep"] * (meGO_LJ["distance"] / meGO_LJ["rc_distance"]) ** 12).clip(
         lower=-20 * meGO_LJ["rep"], upper=-0.05 * meGO_LJ["rep"]
     )[condition]
     meGO_LJ.loc[condition, "learned"] = 1
 
+    #condition = (
+    #    (meGO_LJ["probability"] <= meGO_LJ["limit_rc_att"] * np.maximum(meGO_LJ["rc_probability"], meGO_LJ["rc_threshold"]))
+    #    & (meGO_LJ["probability"] > meGO_LJ["md_threshold"])
+    #    & (meGO_LJ["rc_probability"] <= meGO_LJ["md_threshold"])
+    #)
+    #meGO_LJ.loc[condition, "epsilon"] = -meGO_LJ["rep"] * (
+    #   1.0 + (np.maximum(meGO_LJ["rc_probability"], meGO_LJ["rc_threshold"]) - meGO_LJ["probability"])
+    #)
+    #meGO_LJ.loc[condition, "learned"] = 1
+
     # 1-4 interactions are special and cannot become attractive because they are part of the bonded-interactions
     condition = (meGO_LJ["bond_distance"] < 4) & (meGO_LJ["same_chain"])
     meGO_LJ.loc[condition, "epsilon"] = -meGO_LJ["rep"]
diff --git a/src/multiego/resources/type_definitions.py b/src/multiego/resources/type_definitions.py
index c5748806..300cb54e 100644
--- a/src/multiego/resources/type_definitions.py
+++ b/src/multiego/resources/type_definitions.py
@@ -6,13 +6,15 @@
 mg_HH_c12_rep = 1.2e-8
 mg_ON_c12_rep = 7.5e-7
 mg_NN_c12_rep = 2.5e-5
+
 mg_HO_sigma = 0.169500
-mg_eps_ch3 = 0.15
-mg_eps_HO = 0.15  # hydrogen bond strength
-mg_eps_ch_aromatic = 0.15
+mg_eps_ch3 = 0.10
+mg_eps_HO = 0.13  # hydrogen bond strength
+mg_eps_ch_aromatic = 0.13
 mg_eps_ch2 = 0.10
-mg_eps_pol = 0.12
-mg_eps_ch1 = 0.09
+mg_eps_pol_nz = 0.11
+mg_eps_pol = 0.09
+mg_eps_ch1 = 0.07
 mg_eps_bkbn_O_CB = 0.10
 
 # Dataframe with atom types and associated parameters
@@ -80,7 +82,7 @@
             4.0 * 0.35722**12 * mg_eps_pol,  # "NT", sig=0.35722
             4.0 * 0.31365**12 * mg_eps_pol,  # "NL", sig=0.31365
             4.0 * 0.33411**12 * mg_eps_pol,  # "NR", sig=0.33411
-            4.0 * 0.31365**12 * mg_eps_pol,  # "NZ", sig=0.31365
+            4.0 * 0.31365**12 * mg_eps_pol_nz,  # "NZ", sig=0.31365
             4.0 * 0.31365**12 * mg_eps_pol,  # "NE", sig=0.31365
             4.0 * 0.35812**12 * mg_eps_pol,  # "C",  sig=0.35812
             4.0 * 0.35812**12 * mg_eps_ch_aromatic,  # "CH", sig=0.35812
@@ -106,7 +108,7 @@
             4.0 * 0.35722**6 * mg_eps_pol,  # "NT",
             4.0 * 0.31365**6 * mg_eps_pol,  # "NL",
             4.0 * 0.33411**6 * mg_eps_pol,  # "NR",
-            4.0 * 0.31365**6 * mg_eps_pol,  # "NZ",
+            4.0 * 0.31365**6 * mg_eps_pol_nz,  # "NZ",
             4.0 * 0.31365**6 * mg_eps_pol,  # "NE",
             4.0 * 0.35812**6 * mg_eps_pol,  # "C",
             4.0 * 0.35812**6 * mg_eps_ch_aromatic,  # "CH"
@@ -198,17 +200,29 @@ def lj14_generator(df):
 
 # Special non-local interactions different from basic mg combination rules
 # PROTEIN
-polar_sbtype = ["OA", "OM", "N", "NL", "NT", "NR", "NZ", "NE", "C", "S", "P", "OE", "CR1"]
-hyd_sbtype = ["CH3", "CH3p", "CH2", "CH2r", "CH1"]
+polar_sbtype = ["O", "OA", "OM", "N", "NL", "NT", "NR", "NZ", "NE", "C", "S"]
+hyd_sbtype = ["CH3", "CH2", "CH2r", "CH1", "CAH", "CAH2"]
 special_non_local = [
+    #{
+    #    "atomtypes": (["O", "OM"], ["O", "OM"]),  # charged oxygen-oxygen repulsion
+    #    "interaction": "rep",
+    #    "sigma": None,  # not needed for repulsion
+    #    "epsilon": mg_OO_c12_rep,
+    #},
     {
-        "atomtypes": (["O", "OM"], ["O", "OM"]),  # charged oxygen-oxygen repulsion
+        "atomtypes": (["O"], ["O"]),  # charged oxygen-oxygen repulsion
         "interaction": "rep",
         "sigma": None,  # not needed for repulsion
         "epsilon": mg_OO_c12_rep,
     },
+    #{
+    #    "atomtypes": (["NZ", "NL"], ["NZ", "NL"]),  # charged nitrogen-nitrogen repulsion
+    #    "interaction": "rep",
+    #    "sigma": None,  # not needed for repulsion
+    #    "epsilon": mg_NN_c12_rep,
+    #},
     {
-        "atomtypes": (["NZ", "NL"], ["NZ", "NL"]),  # charged nitrogen-nitrogen repulsion
+        "atomtypes": (["NZ"], ["NZ"]),  # charged nitrogen-nitrogen repulsion
         "interaction": "rep",
         "sigma": None,  # not needed for repulsion
         "epsilon": mg_NN_c12_rep,
@@ -219,18 +233,36 @@ def lj14_generator(df):
         "sigma": None,  # not needed for repulsion
         "epsilon": mg_HH_c12_rep,
     },
-    {
-        "atomtypes": (polar_sbtype, hyd_sbtype),  # polar - hydrophobic repulsion
-        "interaction": "rep",
-        "sigma": None,  # not needed for repulsion
-        "epsilon": None,
-    },  # If None use default rc c12 repulsion
+    #{
+    #    "atomtypes": (polar_sbtype, hyd_sbtype),  # polar - hydrophobic repulsion
+    #    "interaction": "att",
+    #    "sigma": None,  # not needed for repulsion
+    #    "epsilon": 0.09,
+    #},  # If None use default rc c12 repulsion
     {
         "atomtypes": (["O", "OM", "OA"], ["H"]),  # hydrogen bond attraction
         "interaction": "att",
         "sigma": mg_HO_sigma,
         "epsilon": mg_eps_HO,
     },
+    {
+        "atomtypes": (["NL"], ["N", "NT", "NL", "NR", "NZ", "NE", "C", "CH", "CH1", "CAH", "CH2", "CAH2", "CH3", "CH2r", "S"]),  # hydrogen bond attraction
+        "interaction": "rep",
+        "sigma": None,
+        "epsilon": None,
+    },
+    {
+        "atomtypes": (["OM"], ["O", "OM", "C", "CH", "CH1", "CAH", "CH2", "CAH2", "CH3", "CH2r", "S"]),  # hydrogen bond attraction
+        "interaction": "rep",
+        "sigma": None,
+        "epsilon": None,
+    },
+    #{
+    #    "atomtypes": (["CH1", "CAH"], ["CH1", "CAH"]),  # hydrogen bond attraction
+    #    "interaction": "rep",
+    #    "sigma": None,
+    #    "epsilon": None,
+    #},
 ]
 
 # List of amino acids and nucleic acids