add D-COID data set

.gitattributes

@@ -815,3 +815,4 @@ data/dude/part.638.parquet filter=lfs diff=lfs merge=lfs -text
 data/dude/part.713.parquet filter=lfs diff=lfs merge=lfs -text
 data/dude/part.82.parquet filter=lfs diff=lfs merge=lfs -text
 data/dude_predict.parquet filter=lfs diff=lfs merge=lfs -text
+data/dcoid.parquet filter=lfs diff=lfs merge=lfs -text

data/dcoid.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57286fa1878faeb5954050d2c557042e806dd2ad0bf1731a57044d0fff3af785
+size 597990

dcoid.py

@@ -0,0 +1,85 @@
+from mpi4py import MPI
+from mpi4py.futures import MPICommExecutor
+
+from Bio.PDB import PDBParser, PDBIO, Select, PPBuilder
+import warnings
+
+import tempfile
+import os
+import sys
+from rdkit import Chem
+
+import pandas as pd
+
+def is_het(residue):
+    res = residue.id[0]
+    return res != " " and res != "W"
+
+
+class ResidueSelect(Select):
+    def __init__(self, het):
+        self.het = het
+
+    def accept_residue(self, residue):
+        """ Recognition of heteroatoms - Remove water molecules """
+        return (self.het and is_het(residue) or not self.het and not is_het(residue))
+
+def get_complex(fn):
+    try:
+        parser = PDBParser()
+        io = PDBIO()
+        structure = parser.get_structure('complex',fn)
+        io.set_structure(structure)
+
+        with tempfile.NamedTemporaryFile(mode='w',delete=False) as f:
+            name_receptor = f.name
+
+        with tempfile.NamedTemporaryFile(mode='w',delete=False) as f:
+            name_ligand = f.name
+
+        io.save(name_receptor,ResidueSelect(het=False))
+        io.save(name_ligand,ResidueSelect(het=True))
+
+        parser = PDBParser()
+        receptor = parser.get_structure('protein',name_receptor)
+        ppb = PPBuilder()
+        seq = []
+        for pp in ppb.build_peptides(structure):
+            seq.append(str(pp.get_sequence()))
+        seq = ''.join(seq)
+
+        mol = Chem.MolFromPDBFile(name_ligand)
+        smiles = Chem.MolToSmiles(mol)
+
+        os.unlink(name_ligand)
+        os.unlink(name_receptor)
+
+        return seq, smiles
+
+    except Exception as e:
+        print(e)
+        pass
+
+
+if __name__ == '__main__':
+    import glob
+
+    filenames = glob.glob(sys.argv[2])
+    comm = MPI.COMM_WORLD
+    with MPICommExecutor(comm, root=0) as executor:
+        if executor is not None:
+            result = executor.map(get_complex, filenames)
+
+            names = []
+            all_seq = []
+            all_smiles = []
+            for n,r in zip(filenames,result):
+                try:
+                    all_seq.append(r[0])
+                    all_smiles.append(r[1])
+                    names.append(os.path.basename(n))
+                except:
+                    pass
+            df = pd.DataFrame({'name': names, 'seq': all_seq, 'smiles': all_smiles})
+            df.to_parquet(sys.argv[1])
+

dcoid.slurm

@@ -0,0 +1,10 @@
+#!/bin/bash
+#SBATCH -J preprocess_dcoid
+#SBATCH -p batch
+#SBATCH -A STF006
+#SBATCH -t 3:00:00
+#SBATCH -N 20
+##SBATCH --ntasks-per-node=16
+
+srun python dcoid.py data/dcoid_actives.parquet "/gpfs/alpine/world-shared/bip214/binding_affinity/dcoid/actives/*pdb"
+srun python dcoid.py data/dcoid_decoys.parquet "/gpfs/alpine/world-shared/bip214/binding_affinity/dcoid/decoys/*pdb"

dcoid_combine.py

@@ -0,0 +1,12 @@
+import pandas as pd
+
+if __name__ == '__main__':
+    import glob
+
+    df_actives = pd.read_parquet('data/dcoid_actives.parquet')
+    df_decoys = pd.read_parquet('data/dcoid_decoys.parquet')
+
+    df_actives['active'] = pd.Series([True]*len(df_actives))
+    df_decoys['active'] = pd.Series([False]*len(df_decoys))
+
+    pd.concat([df_actives,df_decoys]).to_parquet('data/dcoid.parquet')