Handle scfv (#153)

Co-authored-by: yoel shoshan <[email protected]>
Co-authored-by: yoels <[email protected]>

fusedrug/data/protein/structure/split_scfv_chain.py

@@ -0,0 +1,182 @@
+from jsonargparse import CLI
+from fusedrug.data.protein.structure.structure_io import (
+    load_pdb_chain_features,
+    save_structure_file,
+)
+from typing import Optional, Sequence
+from os.path import isfile, join, dirname, basename
+import os
+import sys
+import subprocess
+import threading
+
+
+def main(
+    *,
+    input_pdb_path: str,
+    input_scfv_chain_id: str,
+    output_pdb_path_extensionless: str,
+    output_heavy_chain_id: Optional[str] = "H",
+    output_light_chain_id: Optional[str] = "L",
+    passthrough_chains: Optional[str] = None,
+    cleanup_temp_files: bool = True,
+) -> None:
+    """
+
+    Takes an input PDB file and allows to split scfv within it to 2 separate chains.
+    This is useful for modifying such PDB to be used in follow up steps that assume such separate chains for heavy and light chain.
+
+    It allows also to "passthrough" additional chains to maintain a "full" PDB.
+
+    Args:
+    input_pdb_path:
+
+    passthrough_chains: optional, will be "pass through", '_' separated if you want multiple
+
+    """
+
+    if passthrough_chains is not None:
+        passthrough_chains = passthrough_chains.split("_")
+
+    loaded_scfv = load_pdb_chain_features(input_pdb_path, input_scfv_chain_id)
+
+    scfv_seq = loaded_scfv["aasequence_str"]
+
+    safety = f"_{os.getpid()}_{threading.get_ident()}"
+
+    scfv_sequence_filename = join(
+        dirname(input_pdb_path),
+        f"sequence_info_{input_scfv_chain_id}_"
+        + basename(input_pdb_path)
+        + safety
+        + ".txt",
+    )
+
+    if not isfile(scfv_sequence_filename):
+        with open(scfv_sequence_filename, "wt") as f:
+            f.write(f">scfv_{input_scfv_chain_id}:...\n{scfv_seq}\n")
+
+    # run anarci:
+    anarci_executable = join(dirname(sys.executable), "ANARCI")
+    if not isfile(anarci_executable):
+        raise Exception(
+            f"ANARCI binary not found in {dirname(sys.executable)}. check installation. You can install it in your env like this: conda install -c bioconda abnumber"
+        )
+
+    anarci_output_filename = join(
+        dirname(input_pdb_path),
+        f"anarci_output_{input_scfv_chain_id}_"
+        + basename(input_pdb_path)
+        + safety
+        + ".txt",
+    )
+
+    if not isfile(anarci_output_filename):
+        subprocess.run(
+            [
+                anarci_executable,
+                "-i",
+                scfv_sequence_filename,
+                "-o",
+                anarci_output_filename,
+            ]
+        )
+    # parse anarci outputs  and obtain separate heavy and light chains:
+    heavy_chain, light_chain = split_heavy_light_chain_from_anarci_output(
+        anarci_output_filename
+    )
+
+    if 0 == len(heavy_chain):
+        raise Exception("ANARCI could not find the heavy chain domain")
+
+    if 0 == len(light_chain):
+        raise Exception("ANARCI could not find the light chain domain")
+
+    # cleanup
+    if cleanup_temp_files:
+        os.remove(scfv_sequence_filename)
+        os.remove(anarci_output_filename)
+
+    heavy_start = scfv_seq.find(heavy_chain)
+    assert heavy_start >= 0
+
+    light_start = scfv_seq.find(light_chain)
+    assert light_start >= 0
+
+    chain_to_atom14 = {
+        output_heavy_chain_id: loaded_scfv["atom14_gt_positions"][
+            heavy_start : heavy_start + len(heavy_chain)
+        ],
+        output_light_chain_id: loaded_scfv["atom14_gt_positions"][
+            light_start : light_start + len(light_chain)
+        ],
+    }
+
+    chain_to_aa_str_seq = {
+        output_heavy_chain_id: loaded_scfv["aasequence_str"][
+            heavy_start : heavy_start + len(heavy_chain)
+        ],
+        output_light_chain_id: loaded_scfv["aasequence_str"][
+            light_start : light_start + len(light_chain)
+        ],
+    }
+
+    chain_to_aa_index_seq = {
+        output_heavy_chain_id: loaded_scfv["aatype"][
+            heavy_start : heavy_start + len(heavy_chain)
+        ],
+        output_light_chain_id: loaded_scfv["aatype"][
+            light_start : light_start + len(light_chain)
+        ],
+    }
+
+    if passthrough_chains is not None:
+        for chain_id in passthrough_chains:
+            curr_loaded_chain_data = load_pdb_chain_features(input_pdb_path, chain_id)
+
+            chain_to_atom14[chain_id] = curr_loaded_chain_data["atom14_gt_positions"]
+            chain_to_aa_str_seq[chain_id] = curr_loaded_chain_data["aasequence_str"]
+            chain_to_aa_index_seq[chain_id] = curr_loaded_chain_data["aatype"]
+
+    saved_files = save_structure_file(
+        output_filename_extensionless=output_pdb_path_extensionless,
+        pdb_id="unknown",
+        chain_to_atom14=chain_to_atom14,
+        chain_to_aa_str_seq=chain_to_aa_str_seq,
+        chain_to_aa_index_seq=chain_to_aa_index_seq,
+        save_cif=False,
+        mask=None,  # TODO: check
+    )
+
+    assert len(saved_files) == 1
+    print(f"saved {saved_files}")
+
+    return saved_files[0]
+
+
+def split_heavy_light_chain_from_anarci_output(filename: str) -> list[Sequence[str]]:
+    # parses ANARCI output on a fasta file of a single heavy and light chain domains
+    heavy_chain = []
+    light_chain = []
+    with open(filename, "r") as file:
+        for line in file:
+            if line.startswith("#"):
+                continue
+            else:
+                parts = line.split()
+                residue = parts[-1]
+                if residue == "-":
+                    continue
+                if line.startswith("H"):
+                    heavy_chain.append(residue)
+                elif line.startswith("L"):
+                    light_chain.append(residue)
+
+    heavy_chain = "".join(heavy_chain)
+    light_chain = "".join(light_chain)
+
+    return heavy_chain, light_chain
+
+
+if __name__ == "__main__":
+    CLI(main, as_positional=False)

fusedrug/data/protein/structure/structure_io.py

@@ -798,10 +798,11 @@ def flexible_save_pdb_file(
         and ((b_factors is None) or isinstance(b_factors, dict))
     )
 
-    assert list(xyz.keys()) == list(sequence.keys())
-    assert list(xyz.keys()) == list(residues_mask.keys())
+    assert set(xyz.keys()) == set(sequence.keys())
+    assert set(xyz.keys()) == set(residues_mask.keys())
+
     if b_factors is not None:
-        assert list(xyz.keys()) == list(b_factors.keys())
+        assert set(xyz.keys()) == set(b_factors.keys())
 
     if only_save_backbone:
         print(