Add FEPPlus benchmark and all pocket pdbs

.gitattributes

@@ -0,0 +1,3 @@
+*.pdb filter=lfs diff=lfs merge=lfs -text
+*.sdf filter=lfs diff=lfs merge=lfs -text
+*.csv filter=lfs diff=lfs merge=lfs -text

fep_benchmark_inputs/README.md

@@ -0,0 +1,23 @@
+# Inputs for running the binding free energy benchmark
+The inputs for the free energy benchmark are divided into two sections:
+* `fep_plus_inputs/` which contains the inputs for the FEP+ software (e.g. `fmp` files) and the scripts required to
+run the benchmark.
+* `structure_inputs/` which contain structural inputs in a format can be read by many computational chemistry platforms
+(i.e. `PDB` and `sdf` files).
+
+This benchmark data set was constructed and validated using FEP+ and the data in the `structure_inputs` was derived from
+the FEP+ `fmp` graph files..
+
+## Arrangment of the input files
+The benchmark data set consists of a number of congeneric series that have been grouped based on the publication where 
+the inputs were based on. As documented in the supporting information of the [accompanying manuscript](https://doi.org/10.26434/chemrxiv-2022-p2vpg),
+many systems were modified from their original source. 
+
+## Metadata
+Details on the congeneric series subset, file naming scheme, protein name, original PDB reference, experimental binding
+affinity dynamic range, number of input structures, and the DOI of the source on the source where the structures were 
+taken from, can be found in the file below:
+* `benchmark_metadata.csv`
+
+The `Number of nodes` column in `benchmark_metadata.csv` refers to the number of ligand structural inputs. Some ligands 
+had multiple structural inputs as they were represented by multiple rotamers or protomer/tautomers.

fep_benchmark_inputs/benchmark_metadata.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cdad4307424417583e8f1e27e195d56e8e5c88cae7354e52f713b16a1f07c07d
+size 8065

fep_benchmark_inputs/fep_plus_inputs/README.md

@@ -0,0 +1,42 @@
+# Benchmark inputs for FEP+
+
+This directory contains the FEP+ input files (`.fmp`s) and submission scripts for running the extended benchmark data
+set. There are 103 `fmp` files for a wide range of targets and congeneric series. The input 
+files are separated into 14 separate directories that reflect the origin of the data sets. Every `fmp` file in this 
+directory forms is part of the benchmark. 
+
+## Contents
+* `apply_all_corrections.sh` Apply post simulation corrections to all output fmp files that require them. This 
+script also collects the names and locations of the output fmp files into text files which facilitate analysis.
+* `post_simulation_corrections.md` A description of the post simulation corrections. These corrections are 
+applied with `apply_all_corrections.sh`.
+* `binding_mode_correction.py` The script to automatically apply binding mode corrections. This is needed for some
+systems.
+* `README.md`. This file. 
+
+## Subset directories
+The names of each directory and the corresponding subset title in the [accompanying manuscript](https://doi.org/10.26434/chemrxiv-2022-p2vpg).
+* `bayer_macrocycles/`: Bayer Macrocycles
+* `charge_annhil/`: FEP+ charge change set
+* `fragments/`: FEP+ fragment data set
+* `gpcrs/`: GPCRs
+* `jacs_set/`: FEP+ R-group set
+* `janssen_bace/`: Janssen BACE1 data sets
+* `macrocycles/`: FEP+ macrocycles
+* `mcs_docking/`: MCS docking sets
+* `merck/`: The public Merck data set
+* `misc/`: Miscellaneous data sets
+* `opls_stress/`: OPLS stress set
+* `opls_ddag`: OPLS private drug discovery data sets
+* `scaffold_hoping/`: FEP+ scaffold hopping
+* `waterset/`: FEP+ buried water set
+
+### Subset directory contents
+Each directory has the following files
+* `subset_metadata.csv`. Metadata on each subsystem, the naming scheme and the PDB each model was based on.  
+* `*.fmp` files. These are FEP+ format maps/perturbation graphs
+* `setup_and_run.sh`. A submission script for all FEP+ maps in the that uses the settings described in the accompanying paper. 
+* ` run_all_corrections.sh`. A script to run all corrections on the fmp files in that directory.
+* Additional files related to post-processing corrections to the results.
+
+

fep_benchmark_inputs/fep_plus_inputs/apply_all_corrections.sh

@@ -0,0 +1,192 @@
+###################
+# Bayer macrocycles
+###################
+echo 'Collecting Bayer macrocycle data'
+
+## Name the outfmp files
+echo  "bayer_macrocycles/ftase_extraligs_custcore_stereo/ftase_extraligs_custcore_stereo_out.fmp" > bayer_macrocycles_outs.txt
+echo "bayer_macrocycles/wagner_brd4/wagner_brd4_out.fmp" >>  bayer_macrocycles_outs.txt 
+
+#######################
+#FEP+ charge change set
+#######################
+echo 'Collecting FEP+ charge set data'
+
+## Apply the corrections
+cd charge_annhil
+./run_all_corrections.sh
+cd ../
+
+## Name the output fmp file locations
+PROTEINS=(cdk2 dlk egfr ephx2 irak4_s2 irak4_s3 itk jak1 jnk1 ptp1b tyk2)
+for P in ${PROTEINS[*]}
+do
+  echo "charge_annhil/${P}/${P}_pkacorr_out.fmp" >> charge_annhil_outs.txt
+done
+echo "charge_annhil/thrombin_whole_map/thrombin_whole_map_out.fmp" >> charge_annhil_outs.txt
+
+############
+# Fragements
+############
+echo 'Collecting fragement set data'
+
+cd fragments
+./run_all_corrections.sh
+cd ../
+
+PROTEINS=(frag_liga_auto frag_mcl1_noweak frag_mup1 frag_p38 hsp90_frag_2rings hsp90_frag_single_ring t4lysozyme_uvt)
+for P in ${PROTEINS[*]}
+do
+    echo "fragments/${P}/${P}_out.fmp" >> fragments_outs.txt
+done
+
+echo "fragments/jak2_set1/jak2_set1_pkacorr_bmcorr_out.fmp" >> fragments_outs.txt
+echo "fragments/jak2_set2_extra/jak2_set2_extra_bmcorr_out.fmp" >> fragments_outs.txt
+
+#######
+# GPCRs
+#######
+echo 'Collecting GPCR data'
+# Apply the corrections
+cd gpcrs
+./run_all_corrections.sh
+cd ../
+
+## Name the files
+echo "gpcrs/p2y1_merged_out.fmp" > gpcrs_outs.txt
+echo "gpcrs/a2a_hip278/a2a_hip278_sbpkacorr_out.fmp" >> gpcrs_outs.txt
+echo "gpcrs/ox2_hip_custcore/ox2_out.fmp" >> gpcrs_outs.txt
+
+#####################################
+# FEP+ R-group perutbation (JACS set)
+#####################################
+echo 'Collecting FEP+ R-group perutbation (JACS set) data'
+cd jacs_set/
+./run_all_corrections.sh
+cd ../
+
+echo "jacs_set/jnk1_manual_flips/jnk1_manual_flips_out_bmcorr_out.fmp" > jacs_set_outs.txt
+echo "jacs_set/mcl1_extra_flips/mcl1_extra_flips_bmcorr_out.fmp" >> jacs_set_outs.txt
+
+PROTEINS=(bace cdk2 jnk1_manual_flips mcl1_extra_flips p38 ptp1b thrombin_core tyk2)
+for P in ${PROTEINS[*]}
+do
+    echo "jacs_set/${P}/${P}_out.fmp" >> jacs_set_outs.txt
+done
+
+################
+# Janssen BACE1
+###############
+echo 'Collecting the Jansseen BACE1 data'
+
+cd janssen_bace 
+./run_all_corrections.sh
+cd ../
+
+echo "janssen_bace/bace_ciordia_retro/bace_ciordia_retro_pkacorr_out.fmp" > janssen_bace_outs.txt
+echo "janssen_bace/bace_keranen_p2/bace_keranen_p2_bmcorr_out.fmp" >> janssen_bace_outs.txt
+echo "janssen_bace/bace_ciordia_prospective/bace_ciordia_prospective_out.fmp" >> janssen_bace_outs.txt 
+echo "janssen_bace/bace_p3_arg368_in/bace_p3_arg368_in_out.fmp" >> janssen_bace_outs.txt
+
+##################
+# FEP+ Macrocycles
+##################
+echo 'Collecting FEP+ macrocycle data'
+PROTEINS=(2B8V_lig24and25_alpha05 2E9P_lig4to7_alpha05 2Q15_lig17to21_alpha05 3RKZ_lig62to70_alpha05 MHT1_lig3_alpha05 ck2_custcore_hotlys hsp90_3hvd_custcore)
+for P in ${PROTEINS[*]}
+do
+    echo "macrocycles/${P}/${P}_out.fmp" >> macrocycle_outs.txt
+done
+
+#############
+# MCS docking
+#############
+echo 'Collecting MCS docking data'
+cd mcs_docking
+./run_all_corrections.sh
+cd ../
+
+echo "mcs_docking/hne/hne_500mM_pkacorr_out.fmp" > mcs_docking_outs.txt
+echo "mcs_docking/renin_customcore/renin_customcore_out.fmp" >> mcs_docking_outs.txt
+
+#######
+# Merck
+######
+echo 'Collecting Merck data'
+cd  merck
+./run_all_corrections.sh
+cd ../
+
+echo 'merck/cdk8_5cei_new_helix_loop_extra/cdk8_symbmcorr_out.fmp' > merck_outs.txt
+echo 'merck/cmet/cmet_out.fmp' >> merck_outs.txt
+echo 'merck/eg5_extraprotomers/eg5_extraprotomers_pkacorr_out.fmp' >> merck_outs.txt
+echo 'merck/hif2a_automap/hif2a_automap_symbmcorr_out.fmp' >> merck_outs.txt
+#echo 'merck/pfkfb3_automap/pfkfb3_automap_symbmcorr_out_fmp' >> merck_outs.txt
+echo 'merck/shp2/shp2_out.fmp' >> merck_outs.txt
+echo 'merck/syk_4puz_fullmap/syk_pkacorr_out.fmp'  >> merck_outs.txt
+echo 'merck/tnks2_fullmap/tnks2_symcorr_pkacorr_out.fmp' >> merck_outs.txt
+
+###############
+# Miscellaneous
+###############
+echo 'Collecting the miscellaneous data sets'
+cd misc
+./run_all_corrections.sh
+cd ../
+
+PROTEINS=(cdk8_koehler galectin3_extra hfaah hiv_prot_ekegren)
+for P in ${PROTEINS[*]}
+do
+    echo "misc/${P}/${P}_out.fmp" >> misc_outs.txt
+done
+echo 'misc/btk_extra_flip/btk_extra_flip_bmcorr_out.fmp' >> misc_outs.txt
+
+######################
+# OPLS drug discovery
+#####################
+echo 'Collecting the OPLS drug discovery data sets'
+PROTEINS=(bathonP_ethers bathonP_thq bathonP_thq_ring hero0 hero1 hero3 hero5 iris lak1 lak2 lak3 orion)
+for P in ${PROTEINS[*]}
+do
+    echo "opls_ddag/${P}/${P}_out.fmp" >> opls_ddag_outs.txt
+done
+
+#############
+# OPLS Stress
+#############
+echo 'Collecting the OPLS stress set data'
+
+PROTEINS=(chk1_set1 chk1_set2 chk1_set3 chk1_set4 chk1_set5 chk1_set6 chk1_set7 cr_bace1 cr_bace2 fxa_set3 fxa_set4 fxa_set5 fxa_set6 hc_bace1 hc_bace2 pb_bace3 fxa_yoshikawa_set)
+for P in ${PROTEINS[*]}
+do
+    echo "opls_stress/${P}/${P}_out.fmp" >> opls_stress_outs.txt
+done
+
+###################
+# Scaffold hoppping
+###################
+echo 'Collecting the scaffold hopping data set'
+
+PROTEINS=(Bace1_4zsp CHK1_3u9n_corehop Era_2q70 Fxa_2ei8 TPSB2_3v7t)
+for P in ${PROTEINS[*]}
+do
+    echo "scaffold_hopping/${P}/${P}_out.fmp" >> scaffold_hopping_outs.txt
+done
+
+##########
+# Waterset
+##########
+echo 'Collecting the water displacement data set'
+
+cd waterset
+./run_all_corrections.sh
+cd ../
+
+PROTEINS=(hsp90_kung brd41_ASH106 taf12 urokinase hsp90_woodhead)
+for P in ${PROTEINS[*]}
+do
+   echo "waterset/${P}/${P}_out.fmp" >> waterset_outs.txt
+done
+echo 'waterset/scyt_dehyd/scyt_dehyd_pkacorr_out.fmp' >> waterset_outs.txt
+echo 'waterset/throm_nozob_hip75/throm_nozob_hip75_bmcorr_out.fmp' >> waterset_outs.txt
+

fep_benchmark_inputs/fep_plus_inputs/bayer_macrocycles/setup_and_run.sh

@@ -0,0 +1,12 @@
+mkdir wagner_brd4
+cp wagner_brd4.fmp wagner_brd4
+cd wagner_brd4
+"${SCHRODINGER}/fep_plus" -HOST bolt_cpu -SUBHOST bolt_gpu -ppj 2 -ffbuilder -ff-host bolt_cpu:50 -time 20000.0 -vacuum -ensemble muVT -seed 1 -JOBNAME wagner_brd4 wagner_brd4.fmp -QARG "-P dev_GPU"
+cd ../
+
+mkdir ftase_extraligs_custcore_stereo
+cp ftase_extraligs_custcore_stereo.fmp ftase_extraligs_custcore_stereo
+cd ftase_extraligs_custcore_stereo
+"${SCHRODINGER}/fep_plus" -HOST bolt_cpu -SUBHOST bolt_gpu -ppj 2 -ffbuilder -ff-host bolt_cpu:50 -time 40000.0 -vacuum -ensemble muVT -core_hopping_lambda_windows 48 -seed 1 -JOBNAME ftase_extraligs_custcore_stereo ftase_extraligs_custcore_stereo.fmp -QARG "-P dev_GPU"
+cd ../
+

fep_benchmark_inputs/fep_plus_inputs/bayer_macrocycles/subset_metadata.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c35be65c292d93a73caf0b697469c10be5ef8f49fa8f04bbc72701090ade620c
+size 335

fep_benchmark_inputs/fep_plus_inputs/binding_mode_correction.py

@@ -0,0 +1,242 @@
+import argparse
+import sys
+from collections import defaultdict
+from typing import List
+
+import numpy as np
+
+from schrodinger.application.desmond.constants import BOLTZMANN
+from schrodinger.application.desmond.measurement import Measurement
+from schrodinger.application.scisol.packages.fep import graph
+from schrodinger.structutils import analyze
+
+
+def clean_bennett_from_map(g: graph.Graph):
+    """
+    Removes the bennett complex and solvent free energy data and replace with the cycle closure correction DDG
+    predictions. This helps to simplify the application of the binding mode correction.
+
+    :param g: An FEP+ graph with output data.
+
+    Note
+    -----
+    The graph is modified in place.
+    """
+    # TODO: As this creates an inconsistency between the DDGs in the graph and the SID, the SID data is removed.
+    # The Jira ticket DESMOND-10777 has been raised to discuss the possibility of updating the SID in the future.
+    for e in g.edges_iter():
+        e.complex_dg = e.ccc_ddg
+        e.solvent_dg = Measurement(0, 0)
+        e.vacuum_dg = None
+        e.complex_sid = None
+        e.solvent_sid = None
+
+
+def get_binding_mode_nodes(g: graph.Graph) -> List[List[graph.Node]]:
+    """
+    Extract the nodes that represent different binding modes of the same
+    ligand. SMILES are used to identify which ligands are the same.
+
+    :param g: A graph that contains multiple binding modes of the ligands, with
+        each binding mode represented as a separate node.
+
+    :return: The nodes of the binding modes as a list of lists. Each entry in
+        the list contains all the nodes of the different binding modes of the
+        same ligand.
+    """
+    # Extract the unique SMILES of each node in the graph and place matching SMILES into the same dict entry
+    mode_nodes_dict = defaultdict(list)
+    for n in g.nodes_iter():
+        smile = analyze.generate_smiles(n.struc, unique=True)
+        mode_nodes_dict[smile].append(n)
+
+    # Return the nodes whose SMILES appear multiple times
+    return [val for val in mode_nodes_dict.values() if len(val) > 1]
+
+
+def calc_mode_corrections(mode_nodes: List[List[graph.Node]],
+                          temperature: float = 300.) -> List[np.ndarray]:
+    """
+    Calculate the binding mode correction for each mode and for each ligand.
+    The presence of other binding modes serves to increase the binding strength
+    of a ligands.
+
+    :param mode_nodes: A list of lists that contains the nodes of the different
+        binding modes of the same ligand.
+    :param temperature: The temperature of the simulation in Kelvin.
+
+    :returns: The binding mode corrections for each ligand. Each correction
+        accounts for the presence of the other nodes of the same ligand.
+    """
+    kT = BOLTZMANN * temperature
+    corrections = []
+    for nodes in mode_nodes:
+        dgs = np.array([n.pred_dg.val for n in nodes])
+        corrs = np.zeros(len(dgs))
+        for i in range(len(dgs)):
+            corrs[i] = -kT * np.log(np.sum(np.exp(-(dgs - dgs[i]) / kT)))
+        corrections.append(corrs)
+
+    return corrections
+
+
+def apply_corrections(mode_nodes: List[List[graph.Node]],
+                      corrections: List[np.ndarray]):
+    """
+    Apply the binding mode correction to each edge that is connected to a node that represents one of at least two
+    binding modes of the same ligand. The edge data are modified in place. and nothing is returned.
+
+    :param mode_nodes: A list of lists that contains the nodes of the different binding modes of the same ligand.
+    :param corrections: The binding mode corrections for each ligand.
+    """
+    for nodes, correction in zip(mode_nodes, corrections):
+        # There is a different correction for each binding mode of the same ligand.
+        for n, corr in zip(nodes, correction):
+            # Apply correction to every edge the node is part of.
+            for e in n.edges:
+                if n == e[0]:
+                    # Deduct the binding mode correction
+                    e.complex_dg -= Measurement(corr, 0)
+                else:
+                    # Add the correction
+                    e.complex_dg += Measurement(corr, 0)
+
+
+def correct_multiple_binding_modes(
+        g: graph.Graph) -> (List[List[graph.Node]], List[np.ndarray]):
+    """
+    Automatically detects different modes, calculates the free energy correction, and applies it to the map. The input
+    map is modified in place.
+
+    :param g: An FEP+ graph that contains multiple binding modes.
+
+    :returns mode_nodes: A list of lists that contains the nodes of the different binding modes of the same ligand.
+    :returns corrections: The binding mode corrections for each ligand.
+    """
+    # Simplify the map by replacing all the bennett DG information with the cycle closure corrected DGs
+    clean_bennett_from_map(g)
+    # Find which ligands have multiple binding modes using canonicalized SMILES:
+    mode_nodes = get_binding_mode_nodes(g)
+    # Calculate the binding mode corrections and apply to the map:
+    corrections = calc_mode_corrections(mode_nodes)
+    apply_corrections(mode_nodes, corrections)
+    g.calc_cycle_closure()
+
+    return mode_nodes, corrections
+
+
+def merge_ligand_nodes(g: graph.Graph, mode_nodes: List[List[graph.Node]],
+                       corrections: List[np.ndarray]):
+    """
+    Merge all nodes of different binding modes into a single one. The most
+    favourable binding mode of each ligand is retained. The graph is modified
+    in place.
+
+    :param g: An FEP+ graph that contains multiple binding modes. Cycle closure
+        must have called for the map for this function to run.
+    :param mode_nodes: A list of lists that contains the nodes of the different
+        binding modes of the same ligand.
+    :param corrections: The binding mode corrections for each ligand.
+    """
+    for n in g.nodes_iter():
+        if n.pred_dg is None:
+            raise ValueError('Cycle closure must be called for the input map.')
+
+    # Determine which nodes to keep and which nodes to discard.
+    # The mode that requires the least correction is the most favourable and it will be retained.
+    to_merge = []
+    to_keep = []
+    for c in corrections:
+        keep_ind = np.where(c == np.max(c))[0][0]
+        to_merge.append(np.where(c != c[keep_ind])[0])
+        to_keep.append(keep_ind)
+
+    # Add the edges of the binding modes into the one that will be retained.
+    nodes_to_delete = []
+    nodes_to_keep = []
+    for nodes, merge_inds, keep_inds in zip(mode_nodes, to_merge, to_keep):
+        for i in merge_inds:
+            # Add the neighbours of the other binding modes to the node that will be kept.
+            for n in nodes[i].neighbors:
+                g.add_edge(nodes[keep_inds], n)
+            nodes_to_delete.append(nodes[i])
+            nodes_to_keep.append(nodes[keep_inds])
+
+    # Looping through the nodes that will be kept, going through the edges, and adding the predicted binding data
+    for nodes in nodes_to_keep:
+        for e in nodes.edges:
+            n_initial, n_final = e.direction
+            e.complex_dg = n_final.pred_dg - n_initial.pred_dg
+            e.solvent_dg = Measurement(0, 0)
+            e.vacuum_dg = None
+
+    # Finally deleting the nodes of the duplicate binding modes.
+    g.remove_nodes_from(nodes_to_delete)
+
+    return nodes_to_keep
+
+
+def main(argv=None):
+    usage = """
+    Let map_out.fmp be an output from FEP+ that has different binding modes of some of the ligands. To
+    calculate the relative binding free between ligands in a way that accounts for the presence of multiple binding
+    modes, type
+
+    $SCHRODINGER/run -FROM scisol binding_mode_correction.py map_out.fmp -o map_bmcorr_out.fmp
+
+    By default, the output 'map_bmcorr_out.fmp' has nodes of the same ligand merged into one. Edges between merged nodes
+    do not have trajectory (fmpdb) information. To not merge the nodes, use the flag '-no-merge', like this
+
+    $SCHRODINGER/run -FROM scisol binding_mode_correction.py map_out.fmp -o map_bmcorr_out.fmp -no-merge
+    """
+    description = """
+    In FEP+, one can run maps that have different binding modes of ligands in different nodes. This script automatically
+    detects multiple binding modes in an FEP+ map and applies the binding mode correction to the calculated relative free
+    energies. When using this script, there are no limitations to the number of ligands and poses, or on the geometry of the map.
+
+    To aid visualization and MUE and RMSE calculation, nodes of the same ligand are automatically merged into a node
+    that represents the most favorable binding mode, although this feature can be switched off.
+
+    When applying the corrections, the Bennett DDGs are replaced with cycle closure corrected DDGs. As this creates an
+    inconsistency between the Bennett DDGs in the map and SID data, the SID data is removed from the map.
+    """
+    parser = argparse.ArgumentParser(usage=usage, description=description)
+    parser.add_argument(
+        'infile',
+        type=str,
+        help="The name of an out fmp file that contains multiple binding modes."
+    )
+    parser.add_argument(
+        '-o',
+        dest='outfile',
+        type=str,
+        help=
+        "The name of the output fmp file that has the binding mode correction for each edge, default='bmcorr_out.fmp'.",
+        default='bmcorr_out.fmp')
+    parser.add_argument(
+        '-no-merge',
+        action='store_true',
+        dest='no_merge',
+        help=
+        'Option to NOT merge nodes of the same binding mode into most favorable, default=False.',
+        default=False)
+    args = parser.parse_args(argv)
+
+    # Load the map that contains ligands with at least 2 binding modes:
+    g = graph.Graph.deserialize(args.infile)
+
+    if g.fep_type != graph.FEP_TYPES.SMALL_MOLECULE:
+        sys.exit("Only small molecule graphs are supported")
+
+    # Detect alternate binding modes and apply the correction:
+    mode_nodes, corrections = correct_multiple_binding_modes(g)
+
+    if not args.no_merge:
+        merge_ligand_nodes(g, mode_nodes, corrections)
+        g.calc_cycle_closure()
+
+    g.write(args.outfile)
+
+
+if __name__ == '__main__':
+    main()

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/cdk2_epikpka.txt

@@ -0,0 +1,4 @@
+pH  7.5
+
+# ligand ligand pka
+39charg 39neu 4.60 

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/cdk2_pka.txt

@@ -0,0 +1,4 @@
+pH  7.5
+
+# ligand ligand pka
+39charg 39neu 3.23

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/dlk_epikpka.txt

@@ -0,0 +1,4 @@
+pH	7.5
+#Lig	pKa
+4neu 4charg     9.87	
+17neu 17charg	8.01

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/dlk_pka.txt

@@ -0,0 +1,4 @@
+pH	7.5
+#Lig	pKa
+4neu 4charg	9.34
+17neu 17charg	6.17

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/egfr_epikpka.txt

@@ -0,0 +1,7 @@
+pH	7.5
+#Lig	pKa
+27neu 27chargR	8.54
+27neu 27chargS  8.54
+28neu 28charg	7.7
+29neu 29charg	9.15
+30neu 30charg	8.74

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/egfr_pka.txt

@@ -0,0 +1,7 @@
+pH	7.5
+#Lig	pKa
+27neu 27chargR	8.25
+27neu 27chargS  8.30
+28neu 28charg	9.48
+29neu 29charg	8.61
+30neu 30charg	8.73

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/ephx2_epikpka.txt

@@ -0,0 +1,4 @@
+pH	7.0
+#Lig	pKa
+19charg 19neu 4.43
+47charg 47neu 4.6

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/ephx2_pka.txt

@@ -0,0 +1,4 @@
+pH	7.0
+#Lig	pKa
+19charg 19neu 3.43
+47charg 47neu 3.25

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/irak4_s2_epikpka.txt

@@ -0,0 +1,4 @@
+pH	7.2
+#Lig	pKa
+5neu 5charg 8.43
+17neu 17charg 7.94

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/irak4_s2_pka.txt

@@ -0,0 +1,4 @@
+pH	7.2
+#Lig	pKa
+5neu 5charg 7.24
+17neu 17charg 7.67

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/irak4_s3_epikpka.txt

@@ -0,0 +1,4 @@
+pH	7.2
+#Lig	pKa
+19neu 19charg 10.33
+28neu 28charg 11.14

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/irak4_s3_pka.txt

@@ -0,0 +1,4 @@
+pH	7.2
+#Lig	pKa
+19neu 19charg 8.65
+28neu 28charg 8.5

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/itk_epikpka.txt

@@ -0,0 +1,3 @@
+pH	7.2
+#Lig	pKa
+1neu 1charg 9.01 

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/itk_pka.txt

@@ -0,0 +1,3 @@
+pH	7.2
+#Lig	pKa
+1neu 1charg 8.91

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/jak1_epikpka.txt

@@ -0,0 +1,4 @@
+pH	7.2
+#Lig	pKa
+16neu 16charg 9.81
+18neu 18charg 8.67

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/jak1_pka.txt

@@ -0,0 +1,4 @@
+pH	7.2
+#Lig	pKa
+16neu 16charg 8.83
+18neu 18charg 6.25

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/jnk1_epikpka.txt

@@ -0,0 +1,4 @@
+pH	7.2
+#Lig	pKa
+19a_neu 19a_charg 8.38 
+19b_charg 19b_neu 4.25

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/jnk1_pka.txt

@@ -0,0 +1,4 @@
+pH	7.2
+#Lig	pKa
+19a_neu 19a_charg 9.46
+19b_charg 19b_neu 2.13

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/ptp1b_epikpka.txt

@@ -0,0 +1,3 @@
+pH	7.0
+#Lig	pKa
+26neu 26charg 11.23 

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/ptp1b_pka.txt

@@ -0,0 +1,3 @@
+pH	7.0
+#Lig	pKa
+26neu 26charg 9.31

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/run_all_corrections.sh

@@ -0,0 +1,8 @@
+PROTEINS=(cdk2 dlk egfr ephx2 irak4_s2 irak4_s3 itk jak1 jnk1 ptp1b tyk2)
+for P in ${PROTEINS[*]}
+do
+    cd "${P}"
+    $SCHRODINGER/run -FROM scisol pka_tautomer_correction.py "${P}"_out.fmp -pka-file ../"${P}"_pka.txt -o "${P}"_pkacorr_out.fmp    
+    cd ../
+done
+

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/setup_and_run.sh

@@ -0,0 +1,10 @@
+PROTEINS=(cdk2 dlk egfr ephx2 irak4_s2 irak4_s3 itk jak1 jnk1 ptp1b thrombin_whole_map tyk2)
+for P in ${PROTEINS[*]}
+do
+    mkdir "${P}"
+    cp ${P}.fmp ${P}
+    cd "${P}"
+    "${SCHRODINGER}/fep_plus" -HOST bolt_cpu -SUBHOST bolt_gpu -ppj 2 -ffbuilder -ff-host bolt_cpu:50 -time 20000.0 -vacuum -ensemble muVT -seed 1 -JOBNAME ${P} ${P}.fmp -QARG "-P dev_GPU"
+    cd ../
+done
+

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/subset_metadata.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:366d46ca0a92d6c91ca1182571cfe156d4c8c4a2024fb78f49fe7e2bbdf02442
+size 1072

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/tyk2_epikpka.txt

@@ -0,0 +1,4 @@
+pH	7.0
+#Lig	pKa
+jmc_3charg jmc_32neu 4.53
+ejm_52neu ejm_52charg 8.80

fep_benchmark_inputs/fep_plus_inputs/charge_annhil/tyk2_pka.txt

@@ -0,0 +1,4 @@
+pH	7.0
+#Lig	pKa
+jmc_3charg jmc_32neu 3.52
+ejm_52neu ejm_52charg 9.36

fep_benchmark_inputs/fep_plus_inputs/fragments/jak2_set1_pka.txt

@@ -0,0 +1,4 @@
+pH  7.0
+
+# ligand ligand percent percent   
+4a 4b 0.5 0.5

fep_benchmark_inputs/fep_plus_inputs/fragments/run_all_corrections.sh

@@ -0,0 +1,11 @@
+cp jak2_set1_pka.txt run_jak2_set1_corrs.sh jak2_set1
+cd jak2_set1
+./run_jak2_set1_corrs.sh
+cd ../
+
+cd jak2_set2_extra
+cp ../run_jak2_set2_corrs.sh .
+./run_jak2_set2_corrs.sh
+rm run_jak2_set2_corrs.sh
+cd ../
+

fep_benchmark_inputs/fep_plus_inputs/fragments/run_jak2_set1_corrs.sh

@@ -0,0 +1,3 @@
+$SCHRODINGER/run -FROM scisol pka_tautomer_correction.py jak2_set1_out.fmp -p jak2_set1_pka.txt -o jak2_set1_pkacorr_out.fmp
+$SCHRODINGER/run -FROM scisol binding_mode_correction.py jak2_set1_pkacorr_out.fmp -o jak2_set1_pkacorr_bmcorr_out.fmp
+

fep_benchmark_inputs/fep_plus_inputs/fragments/run_jak2_set2_corrs.sh

@@ -0,0 +1,2 @@
+$SCHRODINGER/run -FROM scisol binding_mode_correction.py jak2_set2_extra_out.fmp -o jak2_set2_extra_bmcorr_out.fmp
+

fep_benchmark_inputs/fep_plus_inputs/fragments/setup_and_run.sh

@@ -0,0 +1,10 @@
+PROTEINS=(frag_liga_auto frag_mcl1_noweak frag_mup1 frag_p38 hsp90_frag_2rings hsp90_frag_single_ring jak2_set1 jak2_set2_extra t4lysozyme_uvt)
+for P in ${PROTEINS[*]}
+do
+    mkdir "${P}"
+    cp ${P}.fmp ${P}
+    cd "${P}"
+    "${SCHRODINGER}/fep_plus" -HOST bolt_cpu -SUBHOST bolt_gpu -ppj 2 -ffbuilder -ff-host bolt_cpu:50 -time 20000.0 -vacuum -ensemble muVT -seed 1 -JOBNAME ${P} ${P}.fmp -QARG "-P dev_GPU"
+    cd ../
+done
+

fep_benchmark_inputs/fep_plus_inputs/fragments/subset_metadata.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53f0c1b87248e0a7896f0e6f79577f696d027fd186aab6eb5165cb638c329bd7
+size 908

fep_benchmark_inputs/fep_plus_inputs/gpcrs/a2a_epik_pka.txt

@@ -0,0 +1,6 @@
+pH 7.5
+
+4o_4o deprotonated_9.84
+4e D_4eD deprotonated_8.43
+4e U 4eu deprotonated_8.43
+4f_4f deprotonated_10.72

fep_benchmark_inputs/fep_plus_inputs/gpcrs/a2a_symbmcorr.py

@@ -0,0 +1,99 @@
+from schrodinger.application.scisol.packages.fep import graph
+import numpy as np
+import argparse
+
+import sys
+sys.path.append('../')
+import binding_mode_correction as bmc
+
+# Add the symmetry corrections
+degree_of_symm = {'4a':4, '4a flipped':4,
+             '4b D':2,
+             '4b U':2,
+             '4c':4,
+             '4d':4,
+             '4e D':2, '4eD deprotonated':2,
+             '4e U':2, '4eU deprotonated':2,
+             '4f':4, '4f deprotonated':4,
+             '4g':4,
+             '4h':4,
+             '4i':4,
+             '4jR':2,
+             '4jS':2,
+             '4kD':2,
+             '4kU':2,
+             '4lD':2,
+             '4lU':2,
+             '4m D':2,
+             '4m U':2,
+             '4n':4,
+             '4o':4, '4o deprotonated':4,
+             '4q':4,
+             '4r D':2,
+             '4r U':2}
+
+def get_lignames(edge):
+    sep = edge.short_id_title.split('==>')
+    lig1 = sep[0].split(':')[1].strip()
+    lig2 = sep[1].split(':')[1].strip()
+    return lig1, lig2
+
+
+
+def main(argv=None):
+    usage="""
+    Apply the symmetry and binding mode correction to a particular map.
+
+    $SCHRODINGER/run symbmcorr.py map_out.fmp -o map_bmcorr_out.fmp
+    """
+    parser = argparse.ArgumentParser(usage=usage)
+    parser.add_argument('infile',
+                        type=str,
+                        help="The name of an out fmp file that contains multiple binding modes.")
+    parser.add_argument('-o',
+                        dest='outfile',
+                        type=str,
+                        help="If entered, the name of the corrected map, default=None.",
+                        default=None)
+    args = parser.parse_args(argv)
+
+
+    g = graph.Graph.deserialize(args.infile)
+
+    # Remove edges if they don't have predictions:
+    remove_edges = []
+    for e in g.edges_iter():
+        if e.complex_dg is None or e.solvent_dg is None:
+            remove_edges.append(e)
+
+    if len(remove_edges) > 0:
+        print('Removing {} edges as they do not have results'.format(len(remove_edges)))
+        g.remove_edges_from(remove_edges)
+
+    remove_nodes = []
+    for n in g.nodes_iter():
+        if n.pred_dg is None:
+            remove_nodes.append(n)
+
+    if len(remove_nodes) > 0:
+        print('Removing {} nodes as they do not have results'.format(len(remove_nodes)))
+        g.remove_nodes_from(remove_nodes)
+
+    kT =0.596
+    for e in g.edges_iter():
+        l1, l2 = get_lignames(e)
+        total_corr = 0
+        # Symmetry correction
+        total_corr += kT * np.log(degree_of_symm[l1])
+        total_corr -= kT * np.log(degree_of_symm[l2])
+        e.complex_dg += total_corr
+
+    g.calc_cycle_closure()
+
+    mode_nodes, corrections = bmc.correct_multiple_binding_modes(g)
+    bmc.merge_ligand_nodes(g, mode_nodes, corrections)
+    g.calc_cycle_closure()
+    g.write(args.outfile)
+
+if __name__ == '__main__':
+    main()

fep_benchmark_inputs/fep_plus_inputs/gpcrs/run_a2a_corrs.sh

@@ -0,0 +1,3 @@
+$SCHRODINGER/run a2a_symbmcorr.py a2a_hip278/a2a_hip278_out.fmp -o a2a_hip278/a2a_hip278_symbmcorr_out.fmp
+$SCHRODINGER/run -FROM scisol pka_tautomer_correction.py a2a_hip278/a2a_hip278_symbmcorr_out.fmp -o a2a_hip278/a2a_hip278_sbpkacorr_out.fmp -s_ -pka-file a2a_epik_pka.txt
+

fep_benchmark_inputs/fep_plus_inputs/gpcrs/run_all_corrections.sh

@@ -0,0 +1,3 @@
+./run_a2a_corrs.sh 
+$SCHRODINGER/run -FROM scisol merge_graph.py p2y1_meta_sub/p2y1_meta_sub_out.fmp p2y1_ortho_sub/p2y1_ortho_sub_out.fmp -o p2y1_merged_out.fmp -sc
+

fep_benchmark_inputs/fep_plus_inputs/gpcrs/setup_and_run.sh

@@ -0,0 +1,15 @@
+PROTEINS=(a2a_hip278 p2y1_meta_sub p2y1_ortho_sub)
+for P in ${PROTEINS[*]}
+do
+    mkdir "${P}"
+    cp ${P}.fmp ${P}
+    cd "${P}"
+    "${SCHRODINGER}/fep_plus" -HOST bolt_cpu -SUBHOST bolt_gpu -ppj 2 -ffbuilder -ff-host bolt_cpu:50 -time 20000.0 -vacuum -ensemble muVT -seed 1 -JOBNAME ${P} ${P}.fmp -QARG "-P dev_GPU"
+    cd ../
+done
+
+mkdir ox2_hip_custcore
+cp ox2_hip_custcore.fmp ox2_hip_custcore
+cd ox2_hip_custcore
+"${SCHRODINGER}/fep_plus" -HOST bolt_cpu -SUBHOST bolt_gpu -ppj 2 -ffbuilder -ff-host bolt_cpu:50 -time 20000.0 -vacuum -modify_dihe -ensemble muVT -seed 1 -JOBNAME ox2 ox2_hip_custcore.fmp -QARG "-P dev_GPU"
+cd ../

fep_benchmark_inputs/fep_plus_inputs/gpcrs/subset_metadata.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9313a9f142f12eea9556462a94c83da9d1f9c7bc9f229471490419255f8ab5e5
+size 419

fep_benchmark_inputs/fep_plus_inputs/jacs_set/run_all_corrections.sh

@@ -0,0 +1,2 @@
+$SCHRODINGER/run -FROM scisol binding_mode_correction.py jnk1_manual_flips/jnk1_manual_flips_out.fmp -o jnk1_manual_flips/jnk1_manual_flips_out_bmcorr_out.fmp
+$SCHRODINGER/run -FROM scisol binding_mode_correction.py mcl1_extra_flips/mcl1_extra_flips_out.fmp -o mcl1_extra_flips/mcl1_extra_flips_bmcorr_out.fmp

fep_benchmark_inputs/fep_plus_inputs/jacs_set/setup_and_run.sh

@@ -0,0 +1,9 @@
+PROTEINS=(bace cdk2 jnk1_manual_flips mcl1_extra_flips p38 ptp1b thrombin_core tyk2)
+for P in ${PROTEINS[*]}
+do
+    mkdir "${P}"
+    cp ${P}.fmp ${P}
+    cd "${P}"
+    "${SCHRODINGER}/fep_plus" -HOST bolt_cpu -SUBHOST bolt_gpu -ppj 2 -ffbuilder -ff-host bolt_cpu:50 -time 20000.0 -vacuum -ensemble muVT -seed 1 -JOBNAME ${P} ${P}.fmp -QARG "-P dev_GPU"
+    cd ../
+done

fep_benchmark_inputs/fep_plus_inputs/jacs_set/subset_metadata.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:512de45d69255c67939d04439d00582c827e2d195a112abc52eb304ad29de8ed
+size 731

fep_benchmark_inputs/fep_plus_inputs/janssen_bace/bace_ciordia_retro_symbmcorr.py

@@ -0,0 +1,100 @@
+from schrodinger.application.scisol.packages.fep import graph
+import numpy as np
+import argparse
+from itertools import combinations
+
+import sys
+sys.path.append('/home/ross/Work/schrython/')
+import fep_tools as ftools
+import binding_mode_correction as bmc
+
+
+def get_lignames(edge):
+    sep = edge.short_id_title.split('==>')
+    lig1 = sep[0].split(':')[1].strip()
+    lig2 = sep[1].split(':')[1].strip()
+    return lig1, lig2
+
+
+# Add the symmetry corrections. Specific to the map
+degree_of_symm = {'17':2}
+
+def get_error_before_corrs(g):
+    pred_dgs = []
+    exp_dgs = []
+    for n in g.nodes_iter():
+        ligname = n.short_id_title.split(":")[1].strip()
+        if 'flip' not in ligname:
+            pred_dgs.append(n.pred_dg.val)
+            exp_dgs.append(n.exp_dg.val)
+
+    pred_dgs = np.array(pred_dgs)
+    exp_dgs = np.array(exp_dgs)
+
+    pred_dgs += np.mean(exp_dgs) - np.mean(pred_dgs)
+
+    indices = np.arange(len(exp_dgs))
+    pred_pairwise = np.array([pred_dgs[i] - pred_dgs[j] for i, j in combinations(indices, 2)])
+    exp_pairwise = np.array([exp_dgs[i] - exp_dgs[j] for i, j in combinations(indices, 2)])
+
+    sq_diffs = (exp_pairwise - pred_pairwise)**2
+    boot_rmsds = np.zeros(10000)
+    for b in range(10000):
+        samp_inds = np.random.choice(len(sq_diffs), len(sq_diffs))
+        arr_boot = sq_diffs[samp_inds]
+        boot_rmsds[b] = np.sqrt(arr_boot.mean())
+
+    pair_rmsd = np.sqrt(np.mean((exp_pairwise - pred_pairwise)**2))
+    print('Pairwise RMSD without extra rotamers and symmetry correction = {:.2f} +/- {:.2f} kcal/mol'.format(pair_rmsd, np.std(boot_rmsds)))
+
+
+def main(argv=None):
+    usage="""
+    Apply the symmetry and binding mode correction to a particular map.
+
+    $SCHRODINGER/run symbmcorr.py map_out.fmp -o map_bmcorr_out.fmp
+    """
+    parser = argparse.ArgumentParser(usage=usage)
+    parser.add_argument('infile',
+                        type=str,
+                        help="The name of an out fmp file that contains multiple binding modes.")
+    parser.add_argument('-o',
+                        dest='outfile',
+                        type=str,
+                        help="If entered, the name of the corrected map, default=None.",
+                        default=None)
+    args = parser.parse_args(argv)
+
+    g = graph.Graph.deserialize(args.infile)
+
+    ftools.pretty_print_fmp(g)
+    print()
+    print(get_error_before_corrs(g))
+    print()
+
+    # SYMMETRY CORRECTION
+    kT =0.596
+    for e in g.edges_iter():
+        l1, l2 = get_lignames(e)
+        total_corr = 0
+        # Symmetry correction
+        if l1 in degree_of_symm:
+            total_corr += kT * np.log(degree_of_symm[l1])
+        if l2 in degree_of_symm:
+            total_corr -= kT * np.log(degree_of_symm[l2])
+        e.complex_dg += total_corr
+
+    g.calc_cycle_closure()
+
+    # BINDING MODE CORRECTION
+    mode_nodes, corrections = bmc.correct_multiple_binding_modes(g)
+    bmc.merge_ligand_nodes(g, mode_nodes, corrections)
+    g.calc_cycle_closure()
+
+    ftools.pretty_print_fmp(g)
+
+    if args.outfile is not None:
+        g.write(args.outfile)
+
+if __name__ == '__main__':
+    main()

fep_benchmark_inputs/fep_plus_inputs/janssen_bace/bace_retrospective_pka.txt

@@ -0,0 +1,4 @@
+pH  4.5
+
+# ligand ligand  pka
+30 30_protonated 6.45