add rmsd calculation

cwl_adapters/pose_cluster_filter.cwl

@@ -33,6 +33,9 @@ inputs:
     format:
     - edam:format_3814
 
+  filtered_poses:
+    type: string?
+
 outputs:
 
   filtered_poses:

cwl_adapters/rmsd.cwl

@@ -0,0 +1,44 @@
+cwlVersion: v1.2
+
+class: CommandLineTool
+
+baseCommand: ["python3", "/rmsd.py"]
+
+hints:
+  DockerRequirement:
+    dockerPull: mrbrandonwalker/rmsd
+
+inputs:
+
+  predicted_poses:
+    type: File[]
+    inputBinding:
+        itemSeparator: ","
+        position: 1
+        prefix: "--predicted_poses"
+
+  reference_pose:
+    type: File
+    inputBinding:
+        position: 2
+        prefix: "--reference_pose"
+
+  output_json_name:
+    type: string
+    default: "output.json"
+    format:
+    - edam:format_3816
+    inputBinding:
+      position: 3
+      prefix: --output_json_name
+
+  rmsd_output:
+    type: string?
+
+outputs:
+
+  rmsd_output:
+    type: File
+    outputBinding:
+      glob: $(inputs.output_json_name)
+

examples/diffdock/Dockerfile_rmsd

@@ -0,0 +1,15 @@
+
+# docker build -f Dockerfile_rmsd -t mrbrandonwalker/rmsd .
+
+FROM condaforge/mambaforge
+# NOT mambaforge-pypy3 (rdkit is incompatible with pypy)
+
+RUN conda install -c conda-forge rdkit --yes
+
+RUN conda init bash
+
+COPY rmsd.py /
+
+RUN mamba clean --all --yes
+
+ADD Dockerfile_rmsd .

examples/diffdock/rmsd.py

@@ -0,0 +1,29 @@
+import argparse
+from pathlib import Path
+import json
+from typing import Dict
+
+from rdkit import Chem
+from rdkit.Chem import AllChem
+
+parser = argparse.ArgumentParser()
+parser.add_argument('--predicted_poses', help='predicted pose')
+parser.add_argument('--reference_pose', type=str, help='reference pose')
+parser.add_argument('--output_json_name', type=str, help='output json file')
+args = parser.parse_args()
+
+predicted_poses = args.predicted_poses.split(',')
+
+ref_mol = Chem.MolFromMolFile(args.reference_pose)
+pred_mols = [Chem.MolFromMolFile(pose) for pose in predicted_poses]
+predicted_poses = [Path(pose).name for pose in predicted_poses]
+
+
+output_dict: Dict[str, Dict[str, float]] = {}
+for i, pred_name in enumerate(predicted_poses):
+    pred_mol = pred_mols[i]
+    RMSD = AllChem.GetBestRMS(pred_mol, ref_mol)
+    output_dict.update({pred_name: RMSD})
+
+with open(args.output_json_name, 'w', encoding='utf-8') as f:
+    json.dump(output_dict, f)

examples/diffdock/run_diffdock.yml

@@ -68,6 +68,15 @@ steps:
       in:
         predicted_poses: "*output_poses"
         centroid_cutoff: 5 # if centroid of all poses are within cutoff then only keep most confident pose, requires visual inspection
+        filtered_poses: "&filtered_poses"
+
+  - rmsd:
+      scatter: [reference_pose, predicted_poses]
+      scatterMethod: dotproduct
+      in:
+        predicted_poses: "*filtered_poses"
+        reference_pose: "*final_sanitized_sdfs"
+        rmsd_output: "&rmsd"
 
 wic:
   graphviz:
@@ -111,3 +120,7 @@ wic:
       wic:
         graphviz:
           label: Cluster poses
+    (10, rmsd):
+      wic:
+        graphviz:
+          label: Calculate RMSD