Merge pull request #39 from jonathanking/dev

Create a fix for Issue #38.
jonathanking · Nov 3, 2021 · f76150d · f76150d
2 parents ec5b5b6 + a5293c3
commit f76150d
Show file tree

Hide file tree

Showing 2 changed files with 46 additions and 0 deletions.
diff --git a/sidechainnet/utils/load.py b/sidechainnet/utils/load.py
@@ -7,6 +7,7 @@
 import requests
 import tqdm
 
+import sidechainnet as scn
 from sidechainnet.create import format_sidechainnet_path
 from sidechainnet.dataloaders.collate import prepare_dataloaders
 
@@ -261,6 +262,10 @@ def load(casp_version=12,
         local_path = _download_sidechainnet(casp_version, thinning, scn_dir)
 
     scn_dict = _load_dict(local_path)
+
+    # Patch for removing 1GJJ_1_A, see Issue #38
+    scn_dict = scn.utils.manual_adjustment._repair_1GJJ_1_A(scn_dict)
+
     scn_dict = filter_dictionary_by_resolution(scn_dict, threshold=filter_by_resolution)
     if complete_structures_only:
         scn_dict = filter_dictionary_by_missing_residues(scn_dict)

diff --git a/sidechainnet/utils/manual_adjustment.py b/sidechainnet/utils/manual_adjustment.py
@@ -2,6 +2,7 @@
 
 from sidechainnet.structure.build_info import NUM_COORDS_PER_RES
 from sidechainnet.utils.align import binary_mask_to_str
+import sidechainnet as scn
 
 
 def manually_correct_mask(pnid, pn_entry, mask):
@@ -56,3 +57,43 @@ def manually_adjust_data(pnid, sc_entry):
         sc_entry["crd"] = sc_entry["crd"][:-NUM_COORDS_PER_RES * 2]
 
     return sc_entry
+
+
+def _repair_1GJJ_1_A(datadict):
+    """Repair raw Sidechainnet data dictionary by splitting entry 1GJJ_1_A into two.
+
+    The file uploaded to RCSB PDB contains two overlapping domains so a manual adjustment
+    is required. See https://github.com/jonathanking/sidechainnet/issues/38 for more
+    details.
+    """
+    # Locate positions in data arrays
+    found_splits_indices = []
+    for split in scn.DATA_SPLITS:
+        for idx, cur_id in enumerate(datadict[split]["ids"]):
+            if cur_id == '1GJJ_1_A':
+                found_splits_indices.append((split, idx))
+
+    # Carefully split into two entries containing the appropriate data ranges
+    for split, idx in found_splits_indices:
+        for key in datadict[split].keys():
+            if key == 'res':
+                datadict[split][key].append(datadict[split][key][idx])
+            elif key == 'ids':
+                datadict[split][key].append(datadict[split][key][idx] + "2")
+                print(datadict[split][key][-1])
+                datadict[split][key][idx] = datadict[split][key][idx] + "1"
+            elif key == 'ums':
+                datadict[split][key].append(" ".join(
+                    datadict[split][key][idx].split()[110:153]))
+                datadict[split][key][idx] = " ".join(
+                    datadict[split][key][idx].split()[0:50])
+            elif key == 'crd':
+                datadict[split][key].append(
+                    datadict[split][key][idx][110 * NUM_COORDS_PER_RES:153 *
+                                              NUM_COORDS_PER_RES])
+                datadict[split][key][idx] = datadict[split][key][idx][
+                    0 * NUM_COORDS_PER_RES:50 * NUM_COORDS_PER_RES]
+            else:
+                datadict[split][key].append(datadict[split][key][idx][110:153])
+                datadict[split][key][idx] = datadict[split][key][idx][0:50]
+    return datadict