WIP

Author: davidmeijer

README.md

@@ -74,7 +74,8 @@
     or create a dump directly from Docker:
 
     ```bash
-    docker exec -t bionexus-db-1 pg_dump -U bionexus -d bionexus -Fc > /path/on/host/bionexus.dump
+    docker exec -it bionexus-db-1 pg_dump -U postgres -Fc -d mydb -f /tmp/mydb.dump
+    docker cp bionexus-db-1:/tmp/db.dump ./db.dump
     ```
 
 Adminer: http://localhost:8080 (server: db, user: bionexus, db: bionexus)

src/bionexus/db/search.py

@@ -147,14 +147,39 @@ def retro_search_compound(
         from retromol.fingerprint import (
             FingerprintGenerator,
             NameSimilarityConfig,
-            polyketide_family_of
+            polyketide_family_of,
+            polyketide_ancestors_of,
         )
         from retromol.rules import get_path_default_matching_rules
     except ImportError:
         logger.error("RetroMol library is not installed. Please install it to use retro_search.")
         return []
 
-    raise RuntimeError("Please update generator setup to match recent changes in retromol ETL code; centralize generator setup?")
+    COLLAPSE_BY_NAME = {
+        "glycosylation": ["glycosyltransferase"],
+        "methylation": ["methyltransferase"],
+    }
+    
+    def _setup_fingerprint_generator(yaml_path: str) -> FingerprintGenerator:
+        """
+        Setup and return a FingerprintGenerator instance.
+
+        :return: FingerprintGenerator instance
+        """
+        collapse_by_name: list[str] = list(COLLAPSE_BY_NAME.keys())
+        cfg = NameSimilarityConfig(
+            # family_of=polyketide_family_of,
+            # family_repeat_scale=1,
+            ancestors_of=polyketide_ancestors_of,
+            ancestor_repeat_scale=1,
+            symmetric=True,
+        )
+        generator = FingerprintGenerator(
+            matching_rules_yaml=yaml_path,
+            collapse_by_name=collapse_by_name,
+            name_similarity=cfg
+        )
+        return generator
 
     # limit top-k to 500 for performance reasons
     if top_k > 500:
@@ -168,13 +193,14 @@ def retro_search_compound(
     logger.info(f"RetroMol coverage for input compound: {coverage:.2%}")
 
     # Setup generator
-    collapse_by_name = ["glycosylation", "methylation"]
-    cfg = NameSimilarityConfig(family_of=polyketide_family_of, symmetric=True, family_repeat_scale=1)
-    generator = FingerprintGenerator(
-        matching_rules_yaml=get_path_default_matching_rules(),
-        collapse_by_name=collapse_by_name,
-        name_similarity=cfg
-    )
+    # collapse_by_name = ["glycosylation", "methylation"]
+    # cfg = NameSimilarityConfig(family_of=polyketide_family_of, symmetric=True, family_repeat_scale=1)
+    # generator = FingerprintGenerator(
+    #     matching_rules_yaml=get_path_default_matching_rules(),
+    #     collapse_by_name=collapse_by_name,
+    #     name_similarity=cfg
+    # )
+    generator = _setup_fingerprint_generator(get_path_default_matching_rules())
     logger.info(f"Initialized RetroMol FingerprintGenerator: {generator}")
 
     # Calculate fingerprints of shape (N, 512)
@@ -183,6 +209,10 @@ def retro_search_compound(
         logger.warning("No RetroMol fingerprints could be generated for the input compound.")
         return []
 
+    for fp in fps:
+        print(fp.sum())
+    exit("CHK")
+    
     vec_col = "fp_retro_b512_vec_counted" if counted else "fp_retro_b512_vec_binary"
 
     # Cosine distance operator in pgvector is '<=>'; cosine similarity = 1 - distance

src/bionexus/etl/retromol.py

@@ -37,7 +37,8 @@ def _retro_bits_and_vec(
         result_obj = RetroMolResult.from_serialized(result_json)
         generator: FingerprintGenerator = generator
         # Get fingerprint in shape (N, 512); could return multiple fingerprints if there are multiple optimal mappings
-        fps_counted: NDArray[np.int8] | None = generator.fingerprint_from_result(result_obj, num_bits=num_bits, counted=True, kmer_sizes=[1, 2, 3], kmer_weights={1: 2, 2: 4, 3: 8})
+        # fps_counted: NDArray[np.int8] | None = generator.fingerprint_from_result(result_obj, num_bits=num_bits, counted=True, kmer_sizes=[1, 2, 3], kmer_weights={1: 2, 2: 4, 3: 8})
+        fps_counted: NDArray[np.int8] | None = generator.fingerprint_from_result(result_obj, num_bits=num_bits, counted=True, strict=False)
         if fps_counted is None:
             # Can happen when coverage is 0.0
             yield None
@@ -95,6 +96,7 @@ def backfill_retro_fingerprints(
             polyketide_family_of,
             polyketide_ancestors_of,
         )
+        from retromol.rules import get_path_default_matching_rules
     except ImportError:
         logger.error("retromol package not found. Please install retromol.")
         return done
@@ -144,7 +146,8 @@ def _setup_fingerprint_generator(yaml_path: str) -> FingerprintGenerator:
     # Create separate fingerprint generator for each matching ruleset
     generators = {}
     for rid, yaml_path in ruleset_files.items():
-        generators[rid] = _setup_fingerprint_generator(str(yaml_path))
+        # generators[rid] = _setup_fingerprint_generator(str(yaml_path))
+        generators[rid] = _setup_fingerprint_generator(get_path_default_matching_rules())
 
     # Calculate fingerprints
     total_inserted = 0